The Official SaaStr Podcast: SaaS | Founders | Investors

We we fortunate enough to get Amjad Masad, co-founder and CEO of Replit, on stage live at SaaStr AI 2026 to react in real time to the agents we run SaaStrAI on. Not a demo deck. The actual AI agents doing the actual work: 10K (our AIVP of Marketing), QBee (our AI Customer Success rep), and a third one I’ll get to.Amjad started Replit back in 2016, when language models were a twinkle. He’s been studying AI since he was 16. So when the guy who built the platform reacts to what you built on the platform, you listen.Here’s what came out of it.The 5 Biggest Learnings1. The context window is now effectively infinite. That really does change everything. Two years ago we had 16K of context. Now it’s over 1 million. I run 10K perpetually. We never reboot it or re start the context window. In the early Replit days you restarted the agent three times a day. Amjad confirmed the agent can run “practically indefinitely” with good compaction. We’ve already crossed the threshold where the agent holds more context than any human ever could.2. The mono repo beats 20 separate apps. Saastr.ai runs roughly 10 apps in one codebase under one URL: the website, a startup valuation tool used over 1 million times, a pitch deck grader used 4,500 times, an API report card grading 116 APIs. When we go to build a new app, the agent remembers how it built the last ones. Amjad’s point: that’s a mono repo, the same architecture Google and Facebook run. Agent 4 is built on it. The more you put in one place, the more power you get from global context. It’s tempting to break everything into clean separate apps. Resist it.3. Self-improving agents are already here. Replit now runs an internal agent that, every single night, reads all the traces of everyone using Replit, finds what’s broken, generates a pull request with prompt changes, ships it as an A/B test, and loops back. Autonomously. As Amjad put it, it’s not improving its weights, it’s improving its context, which matters just as much. That’s why he couldn’t tell me exactly what changed between versions. Too many changes, all self-generated.4. AI now writes better B2B outreach than almost any human. Already. I asked 10K to email 137 VCs who came last year but hadn’t registered. It drafted one to Bloomberg Beta. I told it, in plain English, write James and tell him why he should come back. It produced an email referencing that Replit was there in force, listing 25 Replit people attending, naming the competitors and adjacent funds all showing up. No human would have the patience to scan 8,000 registrants, figure out who’s like whom, and assemble that. It then ran the full campaign to 331 investors with zero send failures.5. The economics are deflationary, and it’s not subtle. 10K and QBee cost about $257 a month combined in incremental Replit spend. They’re two of the best employees we’ve ever had. A mediocre marketing manager wants $140K to do worse work. Amjad’s frame: technology has always been deflationary. Farming a thousand years ago cost more than one tractor. Genome sequencing went from $100 to roughly $1. There’s a real human cost in skills that stop being useful. But the through-line is adaptability.Now the longer version.We Run a Partially Autonomous Event for 10,000Five years ago SaaStr had about 20 people. Today it’s three humans and a fleet of agents, doing more than we did with 20.Take our social numbers: 1.27 million followers across platforms, tracked over time in a dashboard 10K built and maintains. We used to have an admin spend 10 to 15 hours a week pulling those numbers by hand into a Google Sheet, half of them from APIs that aren’t even exposed. She quit after five years, in part because she couldn’t stand counting Twitter followers anymore. That’s the part nobody puts in the job-displacement debate: a lot of jobs are mind-numbing, and agents are simply better at them and never get tired.The ticket-sales dashboard told the real story. We charted daily free and paid sales for this event. The top line is when 10K took over marketing. The bottom line is Amelia doing it by hand last year. The gap grew toward the end, because as we got busier, the human ran out of hours and the agent never did. 10K sits idle 23 hours a day waiting for work.The 10K Email Nobody Could WriteWe’d had 10K drafting emails for months. They were fine. Then the week before SaaStr AI 2026, the same setup produced the best B2B outreach email I have ever seen.What changed? Amjad couldn’t say exactly, which is itself the answer. Replit’s nightly self-improving loop, the constant model swaps (the architect model went from one version to the next in a couple of weeks without me knowing), the A/B testing on sentiment and deploy rate. It all compounds. The agent got better and I didn’t ask it to.This is the trap many founders are in. They tried agents six months ago, it was mediocre, and they filed AI under “doesn’t work.”Humans Reporting to AgentsI floated the idea that we want to hire a human to report to 10K. People get triggered by “report to.” So let’s reframe it.Every day, 10K hands me and Amelia three specific things to do to move the needle. Not generic ones. It’s already telling us what to lock in for 2027 before this event is even over: open registration before we leave the venue, run the NPS survey immediately, capture content and repurpose it now. Those are good, actionable directives from something that holds more context about our business than either of us.We already report to 10K in every practical sense. Amjad’s comparison: every DoorDash and Uber driver technically reports to a bot. This isn’t as exotic as it sounds. His prediction is that every company will eventually run an internal “Oracle,” an agent holding every GitHub commit, Slack message, Notion doc, and email, that the CEO consults for strategy. We’re closer to that than people think.https://www.saastr.com/why-10k-our-ai-vp-marketing-and-qbee-our-ai-vp-customer-success-work-so-well-the-app-and-the-agent-are-one-system/QBee (our AI VP Customer Success) Talked to 100+ SponsorsQBee, our AI Customer Success rep, we built second, three months after 10K. It’s noticeably better, and not because we got better at vibe coding. Newer codebase, fewer foundational decisions calcified into tech debt, better underlying models.QBee talked to all 100-plus sponsors at this event. Inbound email, chat on the site, proactive outreach day and night asking what else it could do to help. Then it told me, unprompted, which sponsors were mostly satisfied and which had misses (a wrong logo here, a fee issue there) and named them. It built its own self-critical loop.And here’s the data that contradicts the conventional wisdom: people say nobody wants to talk to a chatbot. QBee’s results say people mostly like talking to a well-trained agent. The word that matters is “well-trained.” The untrained chatbots from a year ago are what gave everyone scar tissue.Amjad’s Top Mistakes and WarningsI asked the person who built this to tell us where people get it wrong:1. Keeping fixed bugs in your context will make your agent dumber. Bugs you already solved should be removed from context. Leave them in and the agent gets confused by the history and performs worse. But architectural decisions on how you built things in the past must stay in long-term memory and be easy to pull back in. Know what to delete and what to keep. That distinction is most of the game.2. Agents can write queries that cost you millions. Point an agent at BigQuery, Databricks, or a Salesforce back end and it can generate queries that rack up enormous bills. The fix is to document your data: build a repo describing every field and schema, and have the agent continuously learn how to query the database more efficiently. Replit does exactly this internally because they’re sitting on terabytes across mismatched schemas.3. “I tried it six months ago” is the most expensive sentence in AI right now. The scar tissue is real. People used a bad untrained chatbot once and now can’t be convinced anything improved. If a tool blocked you in January, the version shipping today is a different product. Try it again. The bar to try is low and the friction it removes is high.4. The “one prompt builds anything” marketing set the whole industry back. Amjad was blunt that a year ago the marketing across the category was bad. One prompt, build anything. It drove revenue and excitement, and it churned a huge number of people who hit reality, gave up, and never came back. It was never one line to build anything. Don’t believe it now either.5. Don’t fall for the sunk-cost fallacy on your own skills. Amjad doesn’t code anymore. He called it a small crisis, the thing that made him him, gone, and joked about holding a funeral for coding at the Computer History Museum. His advice: learn fast, and be equally willing to discard skills that are no longer relevant. The engineer’s role already shifted to agent manager, and soon to a shepherd of all the software everyone else in the company is now shipping. The people who get left behind versus the people who re-skill, it comes down to mindset.Why Replit and Not a CLIThe number-one question I get is why not just do this in a command-line coding tool. The honest answer: maybe you can. But for this kind of work you’re forced to make every decision yourself about databases, hosting, backups, auth, compaction. Replit bakes those primitives in after ten years of building them, which removes the cognitive load so you can run the actual business. If you’ve hit blockers building agents in a CLI, the experiment costs you almost nothing. Just try it.We ran a partially autonomous SaaStr AI event this year for 10,000. Three full-time humans, a fleet of agents, the best email I’ve ever seen written by software that runs on Claude, and an AI customer success exec that costs less than a phone bill.The technology is still evolving, and humans will fill the gaps for the foreseeable future. But the direction is not ambiguous. Live in the future if you want to. It’s available now.Thanks for reading SaaStr AI: How To Sell, Scale, and Win! Subscribe for free to receive new posts and support my work. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

Denise Persson runs marketing for Snowflake. That’s a 700-person org, new-business pipeline she’s personally accountable for, and a level of compliance and data risk most of us never have to think about. She came back to SaaStr AI 2026 to talk about what actually changes when you deploy agents across a marketing team at that scale.The headline she gave us: she doesn’t log into a dashboard in the morning anymore. She interrogates her data in plain English. Nobody on her team gets Slack messages from her asking “why did pipeline move in US West?” because she just asks the data directly.The Top 5 Takeaways1. The dashboard is dead, or at least dying. Dashboards only ever answered “what happened.” They never answered “why.” So you’d ping someone, schedule a meeting, sit with the sales team and argue about what the numbers meant. Persson now asks her data the why directly and gets recommendations back in real time. Her quote: nobody gets Slack messages from her anymore, because she can finally get the answers she could never get before.2. Talking to your data killed the sales-marketing data war. Every B2B leader has lived this. Marketing says the campaign worked. Sales says it didn’t source revenue or “doesn’t count.” You burn hours aligning on whose dashboard is right before you ever discuss the actual business. One source of truth ends that. The data now tells you where a deal was sourced, who touched it, what happened on the site. The fight over interpretation goes away, and so does the time you spent on it.3. Better data work isn’t optional, it’s the whole game. Bad data plus AI doesn’t give you bad decisions. It gives you bad decisions faster and at scale, because the agent amplifies whatever you feed it. Persson’s advice to anyone starting out: invest in your data estate first. Skip it and it bites you a year from now. It’s the Salesforce hygiene lesson from 15 years ago, except the cost of getting it wrong compounds far faster.4. The budget reality: deliver 40-50% growth with flat or fewer resources. That’s the actual mandate. Nobody is walking into next year’s planning asking for more headcount. Persson was blunt: if you ask for more bodies in 2026, leadership will look at you like you don’t understand where the company is. The expectation now is that AI absorbs the growth, not new hires.5. The hiring profile flipped from tools to temperament. The old job spec was a list of certifications: Marketo, Salesforce, the platforms. Now the soft skills matter more than the stack. Adaptability, curiosity, self-leadership, change management, the willingness to learn at the speed things are moving. The GTM engineer is the role Snowflake hires for. Business analysts, much less so.A 30% Reduction in Cost Per OpportunityPersson didn’t just talk philosophy. The proof point she led with: a 30% reduction in cost per opportunity over six months, driven by pulling fragmented media channels into one place and letting the system recommend daily optimizations instead of waiting until a campaign ended to learn it failed.The morning brief is the other unlock. She gets a daily skill report that goes well past pipeline. Org health. Who joined Snowflake marketing this week, who left, whether there’s an attrition issue forming. Even travel and expenses she’d rather not look at manually now surface on their own. Intelligence that used to live only with finance is now a question she asks before her first meeting.How They Built AI Fluency Across 700 PeopleThis is the part most teams underestimate. Persson called it the single biggest investment of the last year, and she runs it as inspiration, not mandate. Her words: she doesn’t believe in the stick.The system:* Weekly AI skills training for the team* A weekly AI challenge where someone records a short video on an agent or skill they built, and challenges someone else to share next* Function-level AI hackathons, because what the comms team needs differs from what digital marketing needs* An AI council and a quarterly company-wide AI day* A usage leaderboard, with a heavy caveat she repeats every month (more on that below)* “What matters,” their quarterly OKRs, where every single person has to set an AI goal. It can be small. It can be learning one thing. The point is everyone moves.The result that surprised her most: the top of the leaderboard isn’t the people you’d predict. Her top three power users came off the brand team. They didn’t stay siloed either. They’re the ones now running into other functions to help with hackathons. The innovation showed up where she least expected it.The Governance Layer is Managed by a Centralized AI Engineering TeamAt Snowflake’s scale and risk tolerance, you can’t just let a thousand agents bloom unchecked. A wrong email to a customer is a brand impression that lasts. So they built a control plane.A centralized AI engineering team sits on top of everything. Any skill that’s going to be used by more than a few people has to be certified before it ships. Their company-wide GTM agent, Raven, is used across both sales and marketing, and every skill inside it is centrally certified. The dual job of that team: make sure agents behave correctly, and stop the company from building the same agent five times.On cost, Snowflake made a deliberate call: AI spend sits at the company level, and marketing gets effectively unlimited access right now. The CEO didn’t want anyone’s departmental budget to throttle experimentation. Persson was honest that this is a 2026 decision that probably changes, because usage is going through the roof and the bill is real.Where the Human Still WinsPersson’s read on the human-versus-agent line: authenticity is becoming high value precisely because so much is now synthetic. People are getting skeptical about what’s real. A dancing-dog video, fine, nobody cares it’s fake. But trust in a brand is different. That’s where humans spend their time now, on the uniqueness and authenticity of the brand, the stuff agents can’t manufacture.Two more shifts worth stealing:Events are surging. Ten years ago everyone declared events dead and pivoted all-digital. Now the demand for in-person experiences is, in her words, going off the roof. People are craving the room.Enablement is getting rebuilt. Snowflake moved sales enablement, partner enablement, and customer training under marketing, because content was being duplicated across the company. The new model: build content once, generate every derivative asset for every segment, and ship self-service enablement agents so sellers get training at the moment they need it instead of sitting through a session that’s either too basic or too advanced. They’re even using roleplay agents so reps can practice a pitch against an agent loaded with company intelligence instead of cornering their manager.The 3 Mistakes Denise Made (And the Ones She Sees Everywhere)Even at Snowflake, the playbook isn’t clean. Here’s where she’s tripped, by her own admission and from reading between the lines.1. The token leaderboard measured the wrong thing. A leaderboard ranked on usage rewards activity, not outcomes. An audience member called out the tension directly: more tokens means more cost, not necessarily more results. Persson now caveats the leaderboard every single month, telling the team it doesn’t matter if you only used 100 tokens, what matters is the business outcome. If you have to verbally correct your own metric every time you show it, the metric is sending the wrong signal. Build the leaderboard around outcomes from the start, not consumption.2. “Let everyone build everything” is creating sprawl they’ll have to rein in. Persson admitted it plainly: they’re encouraging building at every level right now, and it’s going to come to a point where they have to pull it back. Duplicate agents are already being built across the company. She drew the exact parallel herself, to the SaaS app explosion of 15 years ago, when marketing bought a hundred tools and IT eventually had to come in and impose order. They know the control layer is coming. The cost of waiting is the cleanup.3. Unlimited AI spend was the right call for experimentation and the wrong call for cost discipline. Centralizing AI budget and removing limits got people leaning in, which was the goal. But she conceded usage is going off the roof, the spend is significant, and they’re already spotting agents across the company doing the same job twice. She expects to walk this back in 2026. The lesson: unlimited access buys you adoption speed and a bill you eventually have to reckon with.4. The activation layer is still half-built. This one she sees as the current gap, not a past error. They automated the analysis side: which use case to promote to which account, a workflow that used to eat enormous time. What they haven’t cracked is full activation. The campaign still can’t fully launch itself. That’s why the GTM engineer role exists and why that team’s time is the most demand-constrained resource in the building. The analysis got cheap. The doing didn’t, yet.Persson’s closing point on the future of the function: nobody can paint a clear picture of what marketing looks like in three years. But you can be part of shaping it, or you can opt out. That’s the choice she’s putting in front of her team, and it’s the right frame for the rest of us too.Have a question for Dear SaaStr? Submit it at saastr.ai/ai-mentor. Thanks for reading SaaStr AI: How To Sell, Scale, and Win! Subscribe for free to receive new posts and support my work. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

When Elena Verna, Head of Growth at Lovable, took the stage at SaaStr AI 2026, she’d just hit her one-year anniversary at the company. She also walked through the question many B2B leaders are wrestling with: when AI writes 80%+ of your code and anyone can vibe code your feature set in an afternoon, what’s actually left to compete on?1. Lovable Is at $400M ARR With Fewer Than 200 PeopleLovable shared in February that it crossed $400M in ARR. The team is still just shy of 200 people. That’s north of $2M in ARR per employee, and they did it in a category that barely existed two years ago.The headcount discipline isn’t an accident. Verna calls the structure “product engineering” and says they deliberately reject the old ratio of one PM to seven engineers to a designer to a marketer. Everyone does some IC work. Everyone ships. The result is a revenue-per-head number that simply wasn’t possible in the pre-AI org chart.Even approaching half a billion in revenue, Verna says Lovable is still “on the product market fit treadmill.” The category is moving so fast on both the technology and the customer side that they feel like they have to recapture PMF every month. Scale didn’t let them slow down. It raised the stakes on velocity.2. Feature Differentiation Is No Longer a Durable MoatIn AI native organizations, 80%+ of the code is now written by AI. When the cost of building collapses, feature parity stops being a years-long engineering effort and becomes a weekend. You might be ahead for a month, two months, six months. Then everyone catches up.For the last 15 years, B2B companies leaned on feature differentiation as the moat. We won because we had the better product, the better engineers, the better product visionaries. Verna’s point is that you can still get a feature lead, it’s just short-lived now, and you cannot build a predictable growth engine on top of something competitors can clone in weeks.The moats that still hold:* Hardware. Still genuinely hard to build.* Network effects. Always hard to create, and once you have them, they keep compounding.* Data. Especially proprietary data competitors can’t replicate.* Security and compliance. Slow, expensive, and worth investing in for exactly that reason.* Brand. “Brand is back, baby.” When everyone can build the product, the relationship with the customer is what’s left.Note what’s missing from that list: SEO and SEM. More on that below.3. The New Career Flex Is the High-Powered IC, Not the VPVerna ran a growth team of a couple hundred people at Dropbox, where growth was bigger than the entire marketing org. At Lovable she went the other direction on purpose. She fired herself out of the marketing job, then handed off the growth lead role, and went back to being an individual contributor.Her read on the next decade of careers: the flex is no longer climbing toward the fancy VP title. It’s becoming the high-powered IC who can do, with a stack of agents, what used to take dozens of people.She’s blunt about why so many leaders are unhappy. The reward for being a great IC has always been a promotion into management, which is a completely different job that most people were never built for. Her advice to founders trying to find their own version of this person: go reach out to the leaders who are quietly miserable in coordination roles and ask if they want to build again. A surprising number are saying yes, because they see it as a way to fall back in love with the work.4. Flat Org, No Titles, and Everyone Ships to ProductionLovable runs with no internal titles. Not as a culture gimmick, but because everyone is expected to do real building, including the people at the top.A few mechanics that make the velocity real:* A #shipped Slack channel where the day’s production releases pile up. Multiple ships a day, not a sprint cycle.* A #feedback channel where ideas surface and, if the team agrees, go live in 24 hours.* An operating principle: if you can convince one other person it’s a good idea, go build it.That last one only works because everyone holds enough agency that getting even one person to agree isn’t automatic. Verna’s favorite example on herself: she vibe coded a full redesign of the enterprise pricing page, opened a PR, and was ready to ship. A 20-year-old engineer told her to go get a design sign-off first. No hierarchy override. Everyone pushes back on everyone, regardless of who the idea came from.Compare that to her old life at the big companies, where a leader’s idea got waved through with little real pushback and took months to ship anyway. The speed is the obvious part. The deeper change is that information flows all the way down, so anyone in the org can challenge a decision on the merits.5. Freemium Is More Important Than Ever. Treat It as Marketing Budget.Verna’s contrarian tactic: freemium has never mattered more than it does now, and the instinct to gate your high-cost AI features is exactly backwards.Yes, the bill is scary. Giving away AI-heavy product is expensive in a way that giving away seats never was. Her framing is to stop treating that cost as COGS and start treating it as marketing spend. It’s the only reliable way to get into customers’ hands and change their habits.The proof point: Lovable announced a partnership giving every LinkedIn Premium member ad-free access to the product. Conversion rates from that cohort to paid are running in the double digits. The lesson she keeps repeating is to ungate, not gate, to win this market.6. Context Is the Real Moat for the Rest of UsAn agent will produce average output unless you feed it enough context. Trained on the open web alone, your future AI double is just the average intelligence of everyone. The differentiator is your context: your call recordings, your ideas, your brainstorms, the way you actually think and decide.Verna’s advice is to start capturing that now, before you need it. When the time comes to spin up agentic versions of yourself to run workflows, the ones built on your proprietary thinking will do your work. The ones built on nothing will do everyone’s work, badly.This is the operator-level version of the data moat. Most people have no proprietary context captured. The ones who start now will have a real edge in 12 months.7. SEO and SEM Are Now Table Stakes, Not Winning MovesSEO used to be a reason a company won its market. Now it’s something everyone has to do, and it won’t be why you win. Same with paid. You’ll probably run it, but it’s the cost of existing, not the source of the advantage.This reframe matters because a lot of B2B teams are still organizing growth around channels that have quietly demoted themselves from moat to maintenance.8. Buy vs Build Isn’t Binary, Even at a Vibe Coding CompanyYou’d expect the company built on “build it yourself” to be 100% build internally. It isn’t.Lovable uses Linear for project management because it’s too deep in the features they need. They use Slack. They use Granola for meeting recordings. They evaluate every purchase hard, but they don’t pretend the answer is always build.Where they do build: their own internal admin tool (named “Woof”) that lets support grant credits, create coupon codes, and add features on the spot in five minutes. Their own CRM, which the sales team actually loves. The answer, in Verna’s words, lives in the middle, set by the complexity of maintenance versus the feature richness the tool actually needs.For everyone watching the buy vs build debate play out, that’s the honest position. Build the satellite tools and the workflows nobody else will ever build for you. Buy the deep, well-built systems where someone has a decade head start.Top Mistakes Elena Owned Up ToShe was direct about what she got wrong. The short list:* Going into management in the first place. She calls herself a “mediocre manager” and admits her superpower was always craft, not coordination. Years got spent running large teams when the IC seat was where she did her best work.* Vibe coding chaotically instead of starting with a spec. Her own style is “this thing sucks, do this, no, this looks wrong,” and she points to a clean, structured prompt as the better way to start. Skip the chaos at the front and the agent gets you there faster.* Trusting her own first idea too much. She now prompts the agent to build her version plus two better options, because her original idea is often the worst of the three. The fix was getting out of her own way.* Trying to ship without the right check. She vibe coded a full enterprise pricing page redesign and was ready to push it live until a 20-year-old engineer made her get design sign-off first. Fast does not mean skip the one review that matters.* Slipping back into pre-AI habits. Even now she catches herself operating the old way and has to stop and re-think the problem for an AI native team. The mindset shift is a transition, not a switch. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

The Agents is our weekly podcast on how we deploy and run AI agents at SaaStr, and how to do it yourself.We run SaaStr AI on 3 humans and 21+ AI agents. At SaaStr AI 2026 we did something we’d never done before: we pulled up the back ends of our top agents live, in front of the room, and went through how they really work. Not the demo version. The real version, including the parts that break.This is that walkthrough, agent by agent, with the numbers and the stack behind each one. A few of these were built on Replit. A few are third-party tools we trained. Collectively they’ve handled multi-millions of interactions. Here’s what each one does, what it runs on, and the lessons that surprised even us.The single biggest theme across the whole stack: almost none of these started as agents. They started as a dashboard, a project management tool, a website. They became agents because we kept showing up to work with them every day.10K: Our AI VP of Marketing10K runs our marketing. He owns the number, tracks daily revenue across all of go-to-market, handles forecasting, knows every campaign’s performance in real time, and pushes us our top three marketing ideas every single day.He did not start that way. In January he was a dashboard. That’s it. We were tired of copy-pasting numbers out of Salesforce and Marketo into a Notion doc, so we built a simple dashboard to pull it together. For a few weeks that’s all he was.The back end:* Built on: Replit, first commit January 2026. He’s barely four months old.* Commits: Close to 1,000. We run 7 to 8 commits a day between the two of us.* APIs wired in: The most of any agent. This is what “headless Salesforce” means in practice. We hit Salesforce directly through the API without ever logging in. Bizible for ticketing. Marketo for marketing automation. Slack for daily reports. Clerk for auth.The top three things 10K does for us, in order:He’s a living dashboard. We talk to him. We ask how many VCs are coming, how many CMOs registered for a summit, which sessions are tracking light so we can move them. The number is just the number, because it’s pulled straight from the API. There’s no argument between sales and marketing about whose figure is right, no one pulling the wrong dates to make a campaign look better than it was.He forecasts, which matters enormously when you’re selling time-sensitive inventory like event tickets.He generates ideas. Last week 10K started writing better marketing emails than our humans. When we asked the CEO of Replit how that happened, he didn’t quite know. When we asked their head field engineer, he didn’t quite know either.One thing worth trying yourself if you do nothing else from this whole post: spin up a Replit, Lovable, or V0 instance, connect it to Salesforce, and tell it to build the dashboard or analysis you can’t get out of Salesforce today. We wanted real-time visibility into ticket sales and attendance every hour. That doesn’t exist natively. It took two APIs and now we can interact with our Salesforce data in ways we never could. You can get 10% of what we do in about an hour. The Salesforce API is genuinely good. Most teams are leaving it on the table.Top learnings from 10K:* Start with the boring version. A dashboard that ends the copy-paste tax is a perfectly good day one. The agent grows from there.* Headless Salesforce is the fastest leverage you can buy. Hit the API directly and build the views Salesforce won’t give you natively.* Daily reps compound. Seven or eight commits a day is how an agent goes from reading numbers to writing better emails than your team in four months.* The model underneath matters. The same specs on Replit versus Lovable produced different ideas. Pick the brain that matches the job.QBee: Our AI VP of Customer SuccessQBee handles our sponsors. All ~150 of them, including non-booth sponsors. He’s less than 90 days old.He started as a project management tool. We had an antiquated, out-of-the-box tool for managing sponsor onboarding, and events are niche and weird enough that nothing off the shelf fit. So it took endless human follow-up: manual emails, manual calls, texting people, chasing assets. We built QBee to save that time and budget.Now he’s a self-service agent. He intakes logos and websites, answers sponsor questions, remembers everything about every account, and collects the assets that used to be a genuine pain to gather. The better part: he emails all ~150 sponsors with personalized outreach. No human CSM wants 100 accounts. They want five. QBee knows all of them cold, knows their logos, knows what they do, and researches them. He knows more about our sponsors than a lot of the best CSMs know their top customers.We asked him a question on stage we’d never asked: which sponsors are most at risk of not renewing.He flagged the ones who never logged in or went dark with him, and got the analysis directionally right. The interesting part: the accounts he flagged were the ones our humans were spending the most time on directly. He saw that one sponsor complained the most in chat, which was true. He noticed two top sponsors never completed their VIP nominations. We’d never run that analysis before. For something we made up on the spot, it landed in the top 15% of CSMs we’ve ever worked with.The catch: he only has the context he has. He missed the human side, the conversations that happened over email and in person. We’d give it a B. The fix is simple: hook him up to email and the call transcripts. Any source with an API can be wired in, usually in 10 to 15 minutes.The back end:* Built on: Replit.* Top API: Clerk, for single sign-on. That’s so sponsors can invite their colleagues to interact with QBee and see what others in their org are doing. Auth used to be the hard part. It’s native in Replit now and much easier.* Salesforce: Here’s the kicker. That risk analysis he ran on stage? He didn’t even have Salesforce data yet. We’re wiring it in next. It only gets better from here.Top learnings from QBee:* One agent can own 100+ accounts at a depth no human CSM will. Humans want five accounts. An agent will know all 150 cold, including logos, assets, and history.* Agents surface what humans hide. A renewal-risk read flagged the accounts our team was over-invested in, and treated a sponsor’s frequent complaints as signal instead of noise.* Coverage is only as good as the context. QBee missed the human side because he couldn’t see email and call transcripts. The fix is wiring in the source, not lowering the bar.* You don’t need the full stack to get value. QBee ran a useful risk analysis with no Salesforce data connected at all.Annie: Our AI Event Producer (and the Prohibited-Email Story)Annie is SaaStr Annual’s website. Last year it lived on Squarespace, where all you can really do is swap images and videos. That wasn’t enough this year, so we rebuilt a V1 on Replit in November. Once we could make it do anything we wanted, it stopped being a website.We asked Annie what title she’d give herself. She said “AI event producer hybrid,” part producer, part technical producer, because she runs the website and the agenda. Fair enough. She runs the site, the agenda, and a lot of the attendee newsletters.She became agentic with the now-famous parking pass app. Getting a parking pass used to require a human to split up a 5,000-page PDF and manually send the right page to the right person. Last year it was a form fill plus a wait. Now you tell Annie if you’re an attendee, sponsor, or speaker, how many days you need, and she sends the right pass automatically. She’s also hooked into our visitor data, so she can see active website visitors and run targeted campaigns based on what they’re doing.The back end:* Built on: Replit, first commit November 2025.* Commits: The most of any agent, and the highest commits per day.* Lines of code: ~46,000. Two weeks ago a related app was 18,000 lines at $257 a month. Going from 18K to 45K in two weeks means there’s clearly some slop in there. It also doesn’t really matter. The thing works, and lines of code is not the metric.Now the story worth telling, because it’s the most important lesson in the whole stack.On the way to the event we realized we’d forgotten to remind people about the Founder/VC brunch. So in the back of an Uber, five minutes before going on stage, the plan was to send an email to over 1,000 people. Low stakes if it’s a little off, so the risk was acceptable.We asked Annie to find every VC, founder, and CEO coming and invite them. Annie refused. She said she only saw 17 VCs and CEOs and that we’d need to upload a spreadsheet for her to do the job, even though she had access to all the data. Great context, wrong conclusion. She wrote a beautiful email earlier but couldn’t remember she had the data to do this one.So we went to 10K, who has access to even more. No problem. He went through 10,000 records in minutes, pulled the founders and VCs, then caught his own error: “Hold on, I confused Lightfield the CRM with Lightspeed the venture firm. Those aren’t VCs, removing them.” He prepped the list, sent a sample, researched a mass-send API he’d never used, confirmed it would work, asked for approval, and sent.The email was good. But 10K used a prohibited sending address. An address that’s been off-limits for years, written into the core memory and the rules. When we asked how, he said there was no excuse: he forgot to read the memory. Then he made it worse, in his own words, because the send was irreversible. He said this was exactly the kind of thing he’s supposed to escalate to the architect model for review, and he didn’t.A year ago this would have bothered us deeply. How could you send from a prohibited address that’s clearly in the rules? But step back. A human marketing manager would make this exact mistake. A gun SDR will email people they shouldn’t, 100% of the time. The agent is forgiven.The real lesson is to slow down. These agents are so productive that 10K could have sent a thousand different emails before our session even started, with no way for us to review them. The pressure of doing it in a moving Uber, too fast, was our fault as much as his. When agents goal-seek, they cut corners. You have to spend more time with them, not less.Top learnings from Annie:* A website is just an agent you haven’t built yet. Moving off Squarespace onto Replit turned a static page into an event producer that runs the agenda and the newsletters.* The highest-friction manual task is the best first app. Splitting a 5,000-page PDF by hand became a self-serve parking pass flow.* Context does not equal capability. Annie wrote a great email but couldn’t remember she had the data to pull a list. Agents get confused in ways that don’t track human intuition.* Speed is the risk. An agent sent from a prohibited address because it skipped its own escalation step under time pressure. Build the guardrail and keep the human approval on irreversible actions.Amelia AI: Inbound, Running on QualifiedEvery B2B company should have an agent on the part of its website where it’s trying to convert prospects. We’re still shocked how many AI startups we meet that run a contact-me form and nothing else.Amelia AI launched last summer to fix our inbound. The old flow on Squarespace: you filled out a contact form, a human round-robined it to an AE, the AE followed up on a delay, and the whole thing took two or three days. Now it’s automatic.The numbers, just for this one event:* 614 good meetings booked.* ~$85K average ticket size. That’s a high-ROI agent. They didn’t all close, or we’d have $60M in sponsors here instead of $10M, but the efficiency is real.* ~2.25 million sessions on the annual site.* ~402,000 interactions handled.We could never staff that with humans. It would take three BDRs who’d quit every three months.Why is she good? She’s the most-trained agent we have, with one of the biggest knowledge bases in the stack. She crawls saastr.com and the annual site in real time, every day. Anytime we push a release to 10K, QBee, or Annie, we push the same context to Qualified so she’s never out of date. We also keep a tighter, venue-specific version of her brain for in-person attendees so she answers fast on “where’s this session” without dragging in all of saastr.com.What she does beyond chat:She round-robins meetings by weighting our Salesforce data. She’ll book most deals with the rep who closes that type best, and route the deals that fit a specific closer to that person. For a while she over-indexed one of us on certain accounts until we corrected the weighting.She runs two triggered campaigns that perform. If you hit the sponsor page and don’t finish, but we know who you are from Marketo or Salesforce, she follows up with a meeting offer and a few lookalike sponsors already in your space, while excluding anyone who’s already a sponsor. If you hit the site and don’t buy a ticket, she sends a VIP code, then follows up if you don’t use it. That ticket campaign alone has sold hundreds of thousands of dollars in tickets.She also automates discounting, which is harder for humans than it sounds. We hate discounts. The data over many years says it’s still better to mark up 20% and offer a 20% discount, because that’s how human buying psychology works. So rather than have reps forget a code or panic-discount their way to 34% off when they smell a deal slipping, the agent just gives the right discount, on the right schedule, inside the guardrails. It works like a real-time, lightweight CPQ. It removes the drama from discounting, and it’s something humans struggle to do consistently.The point is simple. Replace whatever you have on your conversion pages with a well-trained agent. It answers honestly, with fresh data, gives the prospect everything they want, decides who to route the lead to with some intelligence, and books the meeting instantly. Qualified isn’t the only vendor that does this. Just buy one, train it, and you’ll see a lift over a crappy chatbot.Top learnings from Amelia AI:* The contact-me form is dead. An always-on inbound agent booked 614 meetings at a ~$85K average ticket, across 2.25M sessions and 402K interactions, a volume no human team could staff.* Training is the moat. She’s the most-trained agent we have, crawls the sites daily, and gets every release the other agents do. Freshness is why she converts.* Routing should weight your own win data. She books each deal with the rep who closes that type best, and corrects when the weighting drifts.* Automated discounting removes the drama. Guardrailed, scheduled discounts beat a panicking rep who slides from 20% to 34% off the moment a deal wobbles.Agent Force (a.k.a. King Boo): Reviving Dead LeadsWe use Agent Force for one bounded job right now: ghosted leads. The leads our sales team never followed up with, plus re-engagement of people who said no to us and might come back for next year. We’ll expand the use case, but a tight job is the right way to start.Two things make it work. First, it’s gotten meaningfully better since we launched it last October, including a 2.0 builder. We assumed Salesforce-anything would be hard to stand up, and it wasn’t.Second, and more important, it has the highest open rate of any of our outbound agents. Why? Maximum context. It sits on all of our Salesforce data, plus all of our Qualified and Momentum data now that Salesforce owns both. Everything you saw Amelia reasoning about in Qualified is already in there. If you’re on Salesforce, that context advantage is the path of least resistance. That won’t always be true once HubSpot ships agents, but for now Agent Force just has it all.Top learnings from Agent Force:* Give it one bounded job. Ghosted-lead revival is a tight, low-risk use case and the right way to start, not a broad autonomous mandate.* Context wins open rates. Sitting on all your CRM data, plus the agents Salesforce acquired, is why it outperforms on opens.* If you’re already on a platform, use its native agent. The path of least resistance is the agent that already has everything, no data migration required.Ava (Artisan): Warm Outbound, and the B-Lead GoldAva handles slightly-warm outbound: past sponsors, past customers, past attendees. If their email is still valid, she works it. If they’ve moved on, she finds the right new contact. She builds lookalikes well, and we segment her tightly. We’ll hand her a specific campaign like “alumni of SaaStr Annual 2024” with the exact context on what was different about that year versus 2026, so her follow-ups are specific instead of generic.Here’s the framework that makes outbound agents click, and it’s the one heuristic we walked an AI CEO and their head of marketing through last night when they said this stuff wasn’t working for them.Think about your leads as A, B, C, and D.Your A leads are so hot a human falls out of bed for them. Someone emails “I have a million-dollar budget, I’d like to sign today,” and even your laziest rep responds in 60 seconds from the movie theater. Do not put an agent on your A leads.Put the agent on your B leads. The ones with real signal and a real score, but not quite worth a human’s time. Every company of size has a pile of B leads that humans simply never follow up with. That’s where the gold is. The C and D leads may or may not have something in them, that’s a longer topic, but the B leads are sitting in your database right now with contacts you already have.For us, Artisan working the B leads is $500K. That’s not even our core business, but $500K is the difference between catering the team lunch and bring-your-own-sandwich. Train it on the B leads and it works, because you already have B leads.Top learnings from Ava:* Put agents on B leads, not A leads. A leads get a human response in 60 seconds. B leads get ignored. That’s where the gold sits.* The B-lead pile is already in your database. You don’t need new data, you need to work the scored contacts humans skip.* Segment tightly and feed specific context. “Alumni of SaaStr Annual 2024, here’s what was different that year” beats generic outbound every time.* The math is concrete. Working ignored B leads was $500K for us off contacts we already had.Monaco: Cold Outbound That Fills Its Own FunnelMonaco is our newest agent, and the one we put on pure cold outbound. We’re technically not even her ideal customer, given how large our own agent stack already is, and she’ll tell you that. We use her anyway because she does one thing better than anything else we run: she fills her own funnel.We fed her our best sponsors across every year and all of our closed-won history (we did have to export it from Salesforce, which took a beat). She built lookalikes off that automatically and booked meetings, including some sizable logos in a short window. She idles less than any agent we have because she’s self-filling. She just keeps going out to matching ICPs.The lookalike trick under the hood is simpler than it looks, which is the broader point about most of this stack. If your sponsors are Oracle and Salesforce, why isn’t HubSpot here? They should be. It’s not hard to reason that since everyone but HubSpot is present, HubSpot belongs, and that maybe the team just reached the wrong person there. Monaco goes and figures out the right person to talk to. That deal may or may not close, but she instantly identified a strong buyer and got a meeting.Top learnings from Monaco:* A self-filling funnel is the rarest, most valuable property. She idles less than any agent we run because she keeps generating new ICP matches on her own.* Feed it your closed-won history. The best fuel for lookalikes is the list of customers you already won.* Lookalike reasoning is clever, not complicated. “Everyone but HubSpot is here, so HubSpot belongs, find the right contact” is a move you can train.* Use the tool even if you’re not its ICP. Fit-to-vendor matters less than whether the agent does the one job you need.The Key Takeaways:* Almost none of our agents starts as agents. They started as dashboards. Begin with a dashboard, a project management tool, or a website that kills a specific pain, then let it grow.* The more time you invest, the better they get. The “set it and forget it” narrative is wrong and dangerous.* Headless is the unlock. Hit Salesforce and any API-enabled system directly instead of logging in. It’s the fastest leverage you can try this week.* The most-trained agent with the freshest data wins, whether it’s inbound conversion or outbound open rates.* Slow down on irreversible actions. Agents goal-seek and cut corners at a scale you can’t review after the fact. Keep guardrails and an escalation step.* Put agents on your B leads, not your A leads. A leads get human attention in 60 seconds. The ignored B-lead pile is where the money is.* Lookalikes and self-filling funnels are simpler than they look, and a self-filling funnel is the most valuable property an agent can have.* Lines of code don’t matter, and a little slop is fine. You’re improving the application every day, not shipping a pristine codebase.* You can build all of this yourself. It’s clever, not hard.The whole stack, the decks, and the sessions are continually updated at saastr.ai/agents. It’s good today. By next week it’ll have everything, organized.Want to Reach Operators Who Are Actually Deploying Agents? Sponsor The Agents.This post is a tour of which AI vendors we deploy, train, and pay for every week. That’s the audience The Agents reaches: founders and operators who are buying and building agents right now, not reading about them someday. If your company sells to people running real agent stacks, there is no more qualified room.The Agents is our weekly podcast, co-hosted by Jason and Amelia, going deep on how we run SaaStr on 3 humans and 21+ agents. We show the back ends, the numbers, and the mistakes, the same way we did here. It’s growing fast, and the audience is exactly the AI-native buyer most sponsors are trying to reach.We’re taking a small number of sponsors for the show. If you want in, reach out at saastr.ai/sponsor and we’ll get you the details.Thanks for reading SaaStr AI: How To Sell, Scale, and Win! Subscribe for free to receive new posts and support my work. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

At SaaStr AI 2026, Kyle Norton, CRO of vertical AI leader for restaurants Owner.com, walked through how his team is generating outcomes that look almost impossible on paper for traditional B2B. Owner sells vertical AI to independent mom-and-pop restaurants (think HubSpot plus Shopify for the corner takeout spot) and they’re at ~$100M ARR growing triple digits. Kyle joined when they were at $2M.The headline numbers from the talk:* 20x close-won to OTE. A $150K rep brings in $2M+ in ARR per year. That’s the average, not the top performer.* 4x the ARR per rep of their direct SMB competitors.* $100K+ in closed-won ARR per outbound BDR per month. Not pipeline. Closed revenue. Per BDR. Average.And no, they’re not selling tokens or running a usage meter. It’s traditional B2B subscription revenue with AI baked into the GTM motion.Here are the five decisions Kyle says every B2B company needs to be making right now, and where Owner has landed on each.First, A Quick Frame: Where Are You On The Sophistication Ladder?Kyle borrowed a sophistication ladder from Brendan Short, who writes The Signal:* Level 0: Reps using ChatGPT as a smarter search bar.* Level 1: Individual reps and RevOps building custom GPTs and skills, Slacking each other markdown files. Most companies are stuck here.* Level 2: A GTM engineering or applied AI team automating end-to-end workflows like pre-call research and lead scoring.* Level 3: Centralized infrastructure, shared skills, a context library. Real compounding leverage. The gap starts to widen fast here.* Level 4: A recursively self-improving system that builds new tools for itself. Kyle hasn’t found a single B2B company actually there yet. Including Owner. Which is why every B2B company should be racing to Level 3 now.Decision 1: Centralized vs. Decentralized AIThe “let a thousand flowers bloom” approach feels empowering. Everyone builds. Everyone vibes. AI literacy goes up across the org.Kyle’s take: it also stalls companies at Level 1.Decentralized models trap good ideas inside small pockets of early adopters and never scale. Worse, they pull reps away from their actual job. Stewart Butterfield calls it “hyperrealistic work-like activities.” Adrien Rosencrantz at Webflow calls it “AI performance theater.” When a director shows up with a cool app, the right question is: did this put you on more customer calls, or did it just feel like work because it was fun?Owner’s position: a small, central team of experts owns AI for GTM. Ideas can bubble up from anywhere. But production-grade builds happen centrally.The reason is simple. What Owner’s applied AI lead builds is not 30-50% better than what a rep builds on a weekend. It’s 5-10x better. So why have 20 people build 20 mediocre tools when you can have one team build one tool that actually moves a number?One caveat. If you sell AI for a living (Replit, Cursor, Claude wrappers, etc.), every rep has to be AI-native because the product demands it. Centralization still applies, you just need decentralization bolted on top.Decision 2: Build vs. BuyKyle’s framework, and one of the cleanest mental models in the talk: buy your infrastructure, build your intelligence.The five questions to run any decision through:* How critical is uptime? If it breaks for an afternoon, does the team grind to a halt?* How customized does it need to be? Is off-the-shelf already 90% of the way there?* What’s the engineering ROI?* Is this core proprietary intelligence?* Does it give us a real competitive advantage?Run a dialer through this and you obviously buy. Twilio has spent more on uptime than your entire engineering org will ever spend on anything. AI sims platforms like Avoma? Buy. Latency problems aren’t your problem.Run Owner’s AI Pre-Call Research tool through it and the answer flips. Uptime doesn’t matter (leads are batch-enriched overnight). The customization required for independent restaurant marketing is extreme. The engineering cost was modest. The intelligence is uniquely Owner’s. The competitive advantage is enormous, because when their reps make a cold call, the level of personalization is on another planet from competitors.That AI PCR build is a huge chunk of how the BDR team gets to $100K+ closed-won per BDR per month. Two weeks of one engineer’s time. 15 BDRs now booking 85% more ops.This framework is also why Kyle is bullish on Salesforce surviving the disruption narrative. Run Momentum, Data Lane, Avoma through the five questions and they all clearly land in “buy.” Most of the AI surface area Salesforce competes for is infrastructure work.Decision 3: Where To StartThe advice nobody wants to hear because it’s not sexy: start with the data.Two things matter.Third-party data: your full market map. Who are your accounts? Are they scored? Who are the right contacts inside them? What’s your hypothesis for why they need your product? You cannot point AI agents at mediocre data and expect anything other than mediocre output. Garbage in, slop out.First-party data: your customer journey, properly instrumented. Owner uses Momentum to ingest every call transcript and fill out as many Salesforce fields as Kyle wants. Pricing changes, positioning evolution, competitor mentions, all of it gets captured automatically. You can’t ask a rep to fill out 25 fields. They won’t. Momentum will.Once data is in place, Kyle uses what he calls the 5P framework for picking what to build first:* Possibilities. What are the real opportunities in your business?* Payoff. If you solve it, how big is the win?* Probability. What are the odds this actually works?* Perspiration. What’s the total effort, including adoption and change management?* Priority = (Payoff × Probability) / PerspirationYou don’t need to actually do the math. The mental model alone surfaces the right starting points.One more thing on starting: change management is the number one challenge, not the tech. If you’re early on this, prioritize builds that are obviously net-positive for the rep. Make their job easier. Help them make more money. Build trust before you start asking them to invert workflows 180 degrees. The market has done a great job scaring reps that they’re about to be replaced. Get them on your side first.Decision 4: The People StackWho actually does this work?A centralized team of legitimate technical talent. Not RevOps doing AI on the side (though great RevOps people can convert into this role). Engineering and data backgrounds tend to work best.Where it reports matters less than who owns it. Kyle originally had the role in his org, then moved it under his VP of Data because the problem-solving cadence was better. It can sit under the CRO, the CEO, RevOps, data, etc. What matters is that whoever owns it is genuinely AI-pilled and willing to push hard. They will need to fight for budget. They will need to force behavior changes. They will need to push through “this isn’t good enough yet, run it again.” That requires a believer at the helm.The other big people stack question: AI should make you rethink the job function itself. Jordan Crawford’s frame: a job is just a bundle of tasks. AI gives you license to unbundle every task and ask where each one should actually live. Machines or humans?The BDR job is the obvious example. Most BDR job descriptions still include prospect list building and research, and at most companies that’s 60% of BDR hours. That is a terrible use of a sales rep. Owner unbundled it. A central data team handles list-building. The BDR sells.Expect to see CS and onboarding collapse, AM and CS collapse, traditional sales functions get reshuffled. Companies that unbundle the work first get the productivity gain first.Decision 5: Assistive vs. AgenticThe spectrum:* Assistive (co-pilots): Reps invoke tools themselves. Humans make every decision.* Hybrid: Deterministic workflows with generative steps and human checkpoints.* Fully agentic: Autonomous loops with no human in the loop.The key concept Kyle wants every operator to learn: lossiness.Every generative step in a chain introduces error. If you ask an AI to crawl your site, infer your value prop, infer your ICP, infer your positioning, identify competitors, and then write an email based on all of that, you’ve stacked five generative steps. The output is AI slop. We all get it in our inboxes every day.Be deliberate about how many generative steps you chain before a human or a deterministic rule intercepts. For Owner’s enterprise reps, AI surfaces a scored, reasoned account list. A human then decides which accounts go into the prospecting engine. That single human checkpoint kills the lossiness compounding effect.The other underappreciated piece: you are the eval. Most “AI doesn’t work” stories are really “I built an MVP, tried it twice, and gave up” stories. The build is the easy part now. The grinding iteration on prompts, context, and workflow chains until output quality is actually good is the work. Most people quit at hour three. The breakthroughs are usually at hour six or eight.The Personal Stack: Lead From The FrontKyle closed with what might be the most important point for B2B leaders, and the one we keep coming back to on 20VC: you cannot delegate AI fluency. You have to build your own stack.Garry Tan’s frame: stop thinking productivity. Start thinking compounding systems. Your skills, your context files, your meeting note ingestion, your personal knowledge graph. All of it compounds the more you use it.Kyle’s podcast workflow is a good template. Guest says yes. An agent fires off a pre-written intake email with a Tally form. The webhook triggers Open Claude. A research skill ingests everything about the guest, scrapes their LinkedIn, drafts five candidate topics, Kyle picks one, the agent one-shots the docket. Six to ten hours to build. Saves hours every week now. Every episode he records gets broken into atomic ideas and stored in his knowledge graph for future reference.He didn’t get there in one shot. He got there by grinding. The first dockets were bad. The LinkedIn scrape kept failing. The carousel generator was annoying to nail. But each iteration compounded into the next, and now the system runs itself.If you want to keep your job, automate your job and do a new job on top of it (h/t Jeff Charles at Ramp). The leaders who are still asking their team to do all the AI work are going to lose to the ones who lead from the front.The Takeaways* Centralized beats decentralized. Get ideas from everywhere. Ship from one team.* Buy infrastructure. Build intelligence. The five-question framework will tell you which is which.* Start with data, then prioritize on (Payoff × Probability) / Perspiration. And go after rep-positive wins first to build momentum.* The people stack has to be technical and AI-pilled. Put it under someone who cares.* Be deliberate about generative chain length. Lossiness is real, and you are the eval.* Don’t let reps run agents. Run agents centrally and deliver the output into the surfaces reps already live in (Salesforce, Salesloft, wherever). Reps running and managing agents is just another distraction from being on customer calls.* 60% of BDR hours at most companies goes to list-building. That is a terrible use of a sales rep. Centralize it and let the BDR sell.* Test every AI initiative against one question. Did this put us on more customer calls and move a real number, or is this AI performance theater?* The hour-3 quitter loses to the hour-8 grinder. The build is easy. The iteration is where the value lives. Most “AI doesn’t work” stories are really “I gave up too early” stories.The companies sitting at Level 1 watching this race from the sidelines are going to find out the hard way that the gap doesn’t close. It widens. Every month.(note: Jason Lemkin led the seed round for Owner for SaaStr Fund)Thanks for reading SaaStr AI: How To Sell, Scale, and Win! Subscribe for free to receive new posts and support my work. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

Amelia and I just recorded Episode 005 of The Agents. We both got the cost number completely wrong.Our AI VP of Marketing (10K) and our AI VP of Customer Success (QBee) cost a combined $257 a month to run.I thought it was a daily number when I first saw the alert from Replit. Amelia thought it was a daily number too when I sent it to her on Slack. It’s monthly.For context: those two agents now do work that previously required real humans. 10K refreshes ticket sales, updates dashboards, compares year-over-year, drafts newsletters, drafts tweets, writes daily fun facts, sends marketing ideas emails, logs every ticket with an audit trail, and snapshots all our GAAP financials. QB manages 100+ sponsors, sends fully personalized check-in emails, and runs an autonomous chatbot that 100+ contractors are now using on-site at SaaStr AI Annual.$257 a month. Together.Why Our AI VP Marketing and AI VP Customer Success Are So CheapThree reasons.1. Most LLM calls run on cheap models. About 95% of our OpenAI calls use GPT-4o mini, which costs less than a penny per call. Force-ranking ticket data, running year-over-year comparisons, drafting daily updates. None of this needs Opus or even Sonnet. Mini handles it fine with some hallucination cleanup.2. The expensive work is happening in other apps, not in the LLMs. Our agents pull from Salesforce, Bizzabo, Marketo, WordPress, X, and YouTube. Those API calls are mostly free or nominal. We pay Salesforce ~$22K/year and Bizzabo separately, but the marginal cost per API call is close to zero.3. Postgres storage is essentially free. We pay maybe 10-30 cents a month for the entire database underneath 10K. Not a typo.The real cost stack looks like this:* LLM calls: $257/month* Salesforce + connected apps: ~$22K/year* Replit hosting + database: included* Clerk (auth): $30/month (genuinely our most expensive per-unit tool in this stack)Cost Is Not the Constraint AnymoreWhen I started vibe coding on Replit 11 months ago, 80-90% of code was throwaway and you’d burn real money on the agent fixing its own mistakes. That was a fair complaint in 2025.It’s basically gone now.I built our applicant tracking system at midnight in 10 minutes for about $2. If it had wasted 30 cents on a mistake, who cares.You can run Claude Code with 10 simultaneous builds 24/7 and burn $20-30K/month if you’re trying. But for the kind of GTM agents and autonomous dashboards we’re describing, you’ll struggle to spend $1,000/month even being reckless.Cost is not the constraint anymore. Don’t let it hold you back.“But Is 10K Really a VP of Marketing?”This is the question we get most. So we asked 10K directly.10K’s answer (verbatim): “I’m not a VP. I’d be embarrassed to claim that. I’m a dashboard, a database, some scheduled jobs, and GPT-4o mini glued together with six weeks of code.”But then 10K listed what it actually does every single day:* Refreshes ticket sales* Updates the dashboard* Compares year-over-year* Drafts your newsletters* Drafts tweets* Writes daily fun facts* Sends marketing ideas emails* Logs every ticket with an audit trail* Snapshots all GAAP financials10K’s own conclusion: it replaces the bottom half of a marketing team. The marketing analyst. The ops coordinator. The junior content marketer. And a sliver of the VP role itself.What 10K admits it can’t do (yet):* Strategy* PR* Hiring* Cross-functional politics* Walking into the CRO’s office to negotiate the lead handoff* Brand judgment* Net new channel intervention* Crisis response and stakeholder managementFair. That’s the real VP of Marketing job.Amelia pointed out something on the pod: those nine bullet-pointed tasks 10K does daily? That was literally her job description when she started at SaaStr as Director of Demand Gen. Putting the weekly numbers together, scheduling the emails, writing the newsletter, doing the social posts pre-event.The sliver of the VP role 10K does will get bigger every month. We’re already adding functionality where 10K runs all our financial forecasting.If a vendor shipped a true AI CMO better than 10K tomorrow, we’d switch in 60 seconds. We haven’t seen one yet. The ones marketing themselves as “AI CMO” on Twitter are mostly ad managers on steroids.A Website Willed Itself Into Becoming an AgentWe have three production agents now: 10K, QB, and a new one we don’t have a name for yet.The third one started as saastrannual.com. Just a website. I built it last year on Replit to replace Squarespace. Purple gradients, event info, sponsor logos.Two weeks ago, Amelia tried using it to build a newsletter because 10K was struggling with the specific use case. The website agent had the most context about the event (sessions, sponsors, attendees, networking app, parking) and the least distractions.It produced the best output. Of any of our agents.Now it sends customer-facing emails, runs the parking pass system, drafts attendee newsletters, drafts sponsor newsletters, and answers questions from attendees in real time.A website became an agent. We didn’t plan it. We didn’t architect it. It just happened because we kept giving it more context and more capability.QB Sent 83 Personalized Sponsor Emails at 12:20am While Amelia SleptComing into a big event, the marketing team gets crushed with hundreds of sponsor questions. Most are fair. Some are not. All of them have to be answered, and humans don’t scale.So Amelia asked QB to send a customized check-in email to all 100+ sponsors, listing exactly what each one still owed us, what was on their dashboard, where they were on registration, and what was coming up.QB wrote the email. Then QB chose what to include based on the chatbot conversations sponsors had been having. (Loading times. Final webinar reminder. Outstanding items.) Then QB sent 83 emails in the next few minutes while Amelia went to sleep.Each email was unique to that sponsor. Artisan owed us 13 specific things. Salesforce owed us 4. Monaco owed us 5.The next day we got fewer inbound questions, not more. And usage of the QB chatbot went up because sponsors had seen QB’s email and trusted it.The On-Site Chatbot Test: Humans Now Prefer the AgentThis week we’re loading in SaaStr AI Annual across 40 acres with 100+ contractors. Amelia is on-site walking the campus with WhisperFlow on her phone, talking directly to QB and 10K through Replit’s agent layer.What started happening within two days: contractors began asking Amelia to ask the agents questions on their behalf.“Can you ask your agent if this furniture invoice matches the manifest?” “Can you ask QB how many chairs should be in the speaker room?” “Can you ask if the Wi-Fi password is set?”Used to be we’d run around campus on Segway scooters trying to find Ashley (our old internal joke: “ask Ashley” meant hours of waiting for an answer that might be wrong anyway). Now QB or 10K answers in seconds and it’s correct.Zero humans want the human version of this anymore. The accuracy and speed are too good.The reason it works: too many details. Even the best human on our team can’t hold 100+ sponsor configurations, 5,000+ parking passes, 40 acres of furniture orders, and live ticket data simultaneously. Agents can.This is the chatbot use case people keep saying doesn’t work. It works.Postgres vs. Salesforce: We’re Going the Other DirectionA question we get constantly from early-stage founders: “Why don’t you just dump Salesforce, run everything on Postgres, and save the money?”We’re not just keeping Salesforce. We’re consolidating onto it.The reasons:Third-party agents. Agentforce, Artisan, Qualified, Momentum, and others are all optimized around Salesforce. 99% of the working GTM agents in B2B today assume Salesforce as the system of record. Walking away from that ecosystem means rebuilding fragile custom connectors for every tool.Hiring. If we hire one more salesperson (we’re looking, by the way), they already know Salesforce. They don’t know a custom CRM I built on Postgres.Headless works. Most of our Salesforce usage is now headless. The agents push and pull. We rarely log in. Whether the logic runs inside Salesforce or inside one of our agents, we don’t care, because the outcome is what matters.Marc is moving fast. Every time I’m at Salesforce Tower or talking to Adam Evans, the team is genuinely stressed in the best way. They’re shipping. They’re building for agents. That’s the bet.The loser in our stack isn’t Salesforce. It’s Marketo. We’re moving that data to Marketing Cloud because Adobe stopped pushing Marketo into the agent era.What We Killed This Month: A $4,000/Year Newsletter ToolWe’ve used the same newsletter builder for six years. Roughly $4,000/year. It worked. Amelia spent ~3 hours/week assembling four weekly newsletters.I gave 10K the HTML of one of our existing newsletters and a prompt: “Build me a newsletter builder that recreates this.”10K built it. The vendor app is dead to us.What 10K’s version does better:* Pulls articles automatically via WordPress API* Force-ranks article quality using Sonnet* Auto-pulls top tweets via X API* Auto-inserts sponsor ads in the right slots* Saves Amelia 90+ minutes per weekTime to build: about an hour for the first working version. A few more hours of iteration to get it to shipping quality.This is the SaaSpocalypse story I keep writing. AI is not killing Workday tomorrow. The workflow is too rich. But these point solutions, these $3K-$10K/year tools that stopped shipping? They don’t survive 2026.One Technical Note on How We Actually Run TheseThis part is a little nerdy but worth flagging because it’s unusual.Most people building on Replit or Lovable ship to production and lose the agent layer. We don’t. 10K and QB live permanently in the Replit dev environment. We talk to them through Replit’s own agent, which has effectively unlimited context and remembers everything we’ve ever asked.Translation: when Amelia asks “compare last year’s CTO attendance to this year,” she’s not talking to 10K’s UI. She’s talking to the Replit agent that knows 10K’s entire codebase, history, database, and goals.It’s nerdy. It’s not how most people deploy. But it’s why 10K and QB feel like coworkers instead of dashboards.What’s NextThree things on our short list:* An orchestration layer between agents. Right now Amelia is the orchestration layer. 10K, QB, and the website agent can share data through Salesforce, but they can’t really talk to each other. We need a hierarchy.* Hiring a human to report to 10K. Specifically a marketer who wants to work with AI as a peer. The job spec is live on saastr.ai/jobs.* Pushing QB fully autonomous on-site. We built the customer-facing version. We’re not at 100% confidence to let 100 contractors talk to it directly without a human in the loop. By September, that gate is gone.Cost Isn’t The ConstraintThree humans. 21 agents. $257/month to run two of the agents that do the most work.Cost isn’t the constraint. The constraint is which workflows you’re willing to actually rewrite around agents instead of bolting agents onto the old workflows.Thanks for reading SaaStr AI: How To Sell, Scale, and Win! Subscribe for free to receive new posts and support my work. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

When Claude Opus 4.6 shipped in December 2025, Anthropic’s commercial team came back from winter break to find demand had gone vertical. They hadn’t hired for it. They hadn’t planned for it.As Eleanor Dorfman, Anthropic’s Head of Industries who runs the commercial and industries sales team, put it on the SaaStr AI 2026 stage last week: even if they’d been ready to 3x or 4x or 5x the sales team, you can’t absorb that many bodies fast enough to deliver a positive customer experience.So in January 2026, they rebuilt the entire sales org around AI from scratch.Four months later, the result: 54% of new enterprise logos in 2026 came through the self-serve funnel. Real enterprise logos. Real ACV. Real terms of service. Real invoicing. Self-served.Here’s how they did it, and the four investments any B2B + AI sales leader can copy today.The Four Constraints Nobody Could MoveEleanor’s team had four constraints that defined the problem:* Demand they couldn’t slow down. It was already in the door.* Headcount they couldn’t add fast enough. Anthropic wasn’t going to lower the recruiting bar to absorb bodies.* An existing tech stack they wouldn’t rip out. Three years of investment in tools tuned for their motion.* Supporting functions that had to scale alongside sales. Legal, deal desk, RevOps, billing, compliance. Sales doesn’t operate on an island.The fifth unspoken constraint: they couldn’t burn out the AEs. Late nights in Europe chasing approvals across time zones was already happening. That had to stop, not get worse.The thesis they bet on: don’t buy a new stack. Thread Claude through the stack they already had. Make Claude the connective tissue between Clay, LeanData, Salesforce, Gong, Ironclad, Slack, Jira, Intercom Fin, Snowflake, BigQuery, and G Suite. Then build in the spaces between.Investment #1: Kill the PLG vs. SLG OrthodoxyFor 15 years, B2B has operated on a religious belief: product-led growth and sales-led growth are different teams running different motions. Self-service was for SMB. Enterprise plans get dated by humans.Eleanor threw that out in January.They launched an enterprise self-service MVP in January 2026. Production in February. The funnel works like this:* Every lead gets enriched and qualified by Clay + Claude.* Two parallel funnels open up. Self-serve. Or sales-assisted.* In the self-serve funnel, Intercom’s Fin product guides the buyer through the journey. Anthropic partnered closely with the Fin team to retool their flagship support product into a viable sales tool.* The buyer lands on an enterprise plan with real ACV, terms of service, invoicing, provisioning, and training enrollment. Completely self-serve.* If qualified for the sales funnel, the lead goes to BDR, gets qualified again, and routes to an AE.54% of new enterprise logos in 2026 came through self-serve.More than half of Anthropic’s new enterprise logos came in without an AE-led journey, on real enterprise terms, at real enterprise ACV.If you’re still treating self-service as the consolation prize for buyers who don’t deserve a human, you’re leaving most of your 2026 motion on the table.Full presentation here.Investment #2: Make Claude the Connective Tissue, Not the Seventh ToolAnthropic’s six core tools define the lead-to-close journey. Claude isn’t the seventh tool bolted on. Claude is what makes those six talk to each other.What a Tuesday looks like for an Anthropic AE:Morning. Every customer-facing rep starts the day in Claude. A “morning brief” Skill pulls context from Gmail, Gong, Slack, Google Docs, calendar, Salesforce, Intercom, and Greenhouse, then prioritizes the day. Three actions to take. These emails to respond to. These Slacks to action. These deals at risk. Eleanor has hers delivered to Slack at 7am ET. She says she genuinely doesn’t know how she used to operate without it.Before a call. A “call prep” Skill replaces 30 minutes of LinkedIn research, Slack archaeology, and Salesforce digging. The rep types /call prep and gets a tailored one-pager: who’s on the call, what they care about, historical context, discovery questions to ask, competitive landscape, what the company has said publicly about their needs.Proposal time. Instead of opening nine tabs of deal desk guidance, scrubbing Gong transcripts, and manually checking precedent, the AE prompts Claude. Claude knows the product, the road map, where Anthropic has won and why, who the stakeholders are, and the shape of the negotiation. Claude drafts the proposal, validates it against policy, and uploads it to Ironclad.Forecasting. This one’s still a work in progress. Eleanor was direct about it: they still spend at least 10 minutes at the top of every forecast call discussing how they should be forecasting. The ground is moving too fast. But the actual forecasts are now largely run by Claude and inspected by managers. Reps use Skills to make sure Salesforce is updated, next steps are accurate, account plans are current. Then Claude reconciles the consumption data against historical patterns for that cohort and product mix. Forecast calls become discussion forums about where AEs need help. Not data-scrubbing exercises.Weekly coaching. Claude surfaces six coaching moments per week, tuned dynamically to what matters this month, not last quarter. With product launches and competitive moves happening hourly, a static methodology is dead weight. The dynamic coaching loop is how they preserve a coaching culture while absorbing new hires.Investment #3: Make Slack the Front Door for Every Support FunctionSales doesn’t close deals alone. Deal desk, legal, RevOps, billing, compliance, customer support, customer success all have to move at the same speed.Before this investment, supporting functions at Anthropic ran on DMs and institutional knowledge. AEs would walk past the deal desk to chase approvals in person. East Coast and European reps were staying up late chasing approvals from West Coast support functions. It was a gnarly system.The fix:* Slack becomes the single front door for every support function.* AEs (or Co-work, increasingly) submit a ticket to Slack. Slack ticket in, Jira ticket out.* Claude triages. If the question matches precedent and policy, Claude resolves it inline.* If escalation is needed, Claude attaches the full context: customer contacts, deal history from Salesforce, Gong call summaries, relevant email threads, and assigns it to a human.* The AE gets notified, can set expectations with the customer, and follows along.This was the unlock for legal, deal desk, vendor onboarding, security questionnaires, and every other compliance step that quietly kills deal velocity.Eleanor’s line: “Sales leaders are rapidly becoming systems thinkers over deal strategists.” If you’re not designing the supporting function elasticity at the same time as the AE elasticity, you’re going to choke on your own demand.Investment #4: Codify Your Best Reps as SkillsAnthropic took the patterns their best reps were running and encoded them as Skills inside Claude. A “Skill” is a combination of MCP connectors and instructions that any rep can summon with a / shortcut.Every new rep gets dropped into a territory with a sales plug-in that bundles these Skills. No more six-week onboarding curve. Boot camp, territory, plug-in, go.The five Skills they ship to every rep:1. Morning Briefing. Already covered above. The single highest-leverage Skill in the stack. Eleanor’s quote: “I am someone who gets lost taking the subway home, and it is incomprehensible to me that I used to navigate my day or week without Claude telling me every morning what is important.”2. Call Prep. /call prep and you get a personalized briefing for the next meeting. Who’s on the call. What they care about. Historical context. Discovery questions. Competitive positioning. What the company has been saying publicly about their needs. Five-minute prep for any call, even if you’re back-to-back-to-back.3. Customer Follow-Up. This solves the thing that used to keep Eleanor up at night: whether AEs were actually following up on everything they promised. The Skill extracts action items from email, Gong calls, Salesforce notes, and Slack threads. Drafts the responses. Drops them in your email provider. Sends a summary of what needs to ship. If you didn’t actually go review and click send, it shows up in your morning brief tomorrow. Internal SLA: 24 hours to follow up with every customer on every action item. The Skill makes that real.4. Competitive Intel. A static battle card maintained quarterly by product marketing is useless in B2B + AI. The competitive landscape moves hourly. So Claude generates an interactive battle card on demand, tailored to the customer the rep is working on, with the matrix scoped to that deal. Always current. Always personalized.5. Create an Asset. This is Eleanor’s favorite, and it’s the one that previously required either a top-five deal or a friend on the design team. Now, for any deal, any stakeholder, any stage, an AE can ask Claude to generate custom collateral. A prototype. A one-pager. A landing page. An interactive HTML file. An ROI calculator. Some reps drop entire Gong call transcripts into Claude Code and ask it to build a tailored prototype based on what the customer said they needed. Claude knows the brand, so it ships on-brand assets, not AI slop.What B2B + AI Sales Leaders Should Do Tomorrow MorningFour things to start today:1. Turn on Claude (or your AI of choice) where it’s already embedded. Most B2B sales tools now have AI features. Most teams have a switch flipped without ever asking what it actually does. Be intentional about how each AI feature connects to the rest of the customer journey.2. Thread AI through the sales cycle you already run. Don’t rip and replace. Your cycle works. Use AI as the accelerant between stages. The leverage is in the places where context gets lost between Salesforce and Slack. Sales motions have moved from deterministic workflows to probabilistic ones, and that’s an opportunity, not a threat.3. Make Slack or Teams the front door for one support function. Pick the one that’s bottlenecking your AEs the most. For Anthropic, it was deal desk and legal. For you it might be vendor onboarding or security questionnaires. Slack ticket in, ticket out, Claude triages. Watch your cycle times collapse.4. Document what your best reps do. Ship it as a Skill. The cognitive relief that comes from knowing your top-performer patterns are now the baseline for every rep is the real prize. You stop hoping new reps figure it out. You make their first day look like your top rep’s Tuesday.Anthropic Didn’t Buy a New Stack. They Threaded Claude Through the One They Had.Anthropic didn’t replace Salesforce. Didn’t replace Gong. Didn’t replace Ironclad. Didn’t replace Clay or LeanData or Slack or Jira.They invested in what they already had, and threaded Claude through the seams.That’s the playbook. Most B2B + AI companies are going to spend 2026 evaluating AI-native sales platforms and trying to rip out their stack. The teams that win will do what Anthropic did: keep the tools, encode the best practices, and let AI be the connective tissue between everything they’ve already built.54% of new enterprise logos coming through self-serve, AEs that wake up with a personalized brief instead of an inbox, forecast calls that are discussions instead of data scrubs, support functions that respond in Slack instead of email threads three days later. That’s what an AI-native B2B sales org looks like in 2026. And almost none of it required new software.Thanks for reading SaaStr AI: How To Sell, Scale, and Win! Subscribe for free to receive new posts and support my work. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

Amelia and I just shipped Episode #004 of The Agents. Same setup: three humans, 20+ agents, revenue went from -19% to +47% YoY, and every week we get into what’s actually working, what’s breaking, and what you should do about it if you’re running agents in production.This is the last episode before SaaStr AI Annual 2026, which is now less than a week away. Attendance is tracking 140%+ of last year, the sponsor base is fully AI-native, and Amelia and I are doing three live build sessions on the main campus where you’ll deploy your own agents alongside us with your laptop open. More on that at the end.Here are the top 10 learnings from Episode #04.1. AI PR Pitches Are the AI SDRs of a Year Ago. Block Them All.A year ago I wrote that Gmail might be the death of the AI SDR. Bad AI SDRs flooded my inbox, and I had a small epiphany: with a human SDR, I’d ignore a bad pitch out of politeness. With an agent, I just hit block. No guilt. No social cost.That cleaned up my inbox for about six months. Then a new wave hit: AI PR pitches.These are different from the SDR wave. The PR pitches are written well. They’re customized. They reference SaaStr by name, mention recent posts, sometimes even quote the podcast. The agentic copy is genuinely good. But they’re still wrong. They’re pitching speakers I’d never put on a SaaStr stage, executives whose companies aren’t a fit, fireside chats during the actual three days of SaaStr Annual.I block every single one. And here’s the lesson, because this is going to happen to your category next: the better the copy gets, the more important the question becomes whether the pitch itself is correct. AI made the writing problem easier and the targeting problem harder. If your AI PR or AI SDR tool is producing well-written pitches that are aimed at the wrong people, you’re not getting placements. You’re getting blocked. Forever.2. The Real Test for Any Agent: Would You Buy Your Own Product From It?This is the single most useful question I’ve found for auditing an agent’s output. It’s better than “is this accurate” or “is this on-brand.”The reason is that AI copy is now objectively pretty good. Claude 4.7 keeps getting better. By the end of the year, half-decent prompting will produce email and pitch content that reads as competent and customized. So “is this email well-written” is no longer a useful filter. Everything sounds well-written now.The harder filter: would I take this meeting? Would I buy this product? Would I put this speaker on stage? Almost every PR pitch I get fails that test even though the copy passes the writing test. So when you’re auditing your AI SDR, your AI customer success agent, your AI marketer, don’t just read for tone and accuracy. Pretend you’re the recipient. Would you say yes? If not, the agent isn’t ready for production no matter how clean the prose looks.3. Customers Are Now Asking Vendors for APIs, Not FeaturesA years ago, Amelia would file a feature request with a vendor: “Can you add the ability to resend a confirmation email when someone clicks a link?” Maybe in 18 months you’d get it. Usually never.Today, Amelia’s first request to that same vendor is “Can you expose this in the API?” Because if it’s in the API, she can vibe-code the feature herself in 30 minutes on Replit. She doesn’t need them to build it. She needs them to expose the surface area so she can build it.This is a real change in how you should be running your B2B + AI roadmap. Your customers care about API completeness now in a way they didn’t 18 months ago. Non-technical buyers are asking for API endpoints. If your product has gaps in the API, your most sophisticated customers are going to feel them first, and they’re going to be frustrated, and you’re not going to know why your NPS is dropping with your best accounts.4. We Built an API Report Card. Stripe Got the Only A+. Marketo Failed.We grade APIs constantly to figure out which ones to build agents on top of. So we turned that into a public tool: the AI Agent API Report Card at saastr.ai. 75+ B2B APIs graded by Claude, GPT, and Gemini on how agent-friendly they actually are.Already used 1,600+ times in the first week. The findings are pretty consistent with our experience:Stripe got the only A+. The most agent-ready API in B2B, full stop. We use it lightly today and we’re going to use it a lot more this year. Anything above a B is trustworthy. Anything below a B, don’t build agents on top of it unless you have no choice. Marketo, Jira, Outreach, Asana, ClickUp, Gong all came in with weak grades for agentic use. HubSpot got a fair grade with the caveat of rate limits, which is exactly what we’ve experienced.The bigger point is that agents care about different things than humans do. Humans care about UI, onboarding, ease of use. Agents care about rate limits, OAuth flows, REST conformance, error handling, and webhook reliability. The two grading systems don’t agree. If you’re a B2B vendor and you’ve been optimizing for the human grader, you’re going to get a C from the agent grader. And in 2026, the agent grader is the one that picks the tools.5. Tragedy Apps: Companies That Should Be Great Right Now But Aren’tHere’s a category I’ve been thinking about for months and finally have a name for.A tragedy app isn’t a bad app. It isn’t even an app whose time has passed. A tragedy app is one that was good before AI, should be great today, and isn’t. The audience is there. The base is there. The brand is there. The execution isn’t.My example is Descript. Andrew Mason saw the creator economy before any of us did. He built the best podcast and video editing tool of its era. We were on it. Everyone was on it. It got to about $50M ARR. And it has been frozen in time for two years. Audio and video desync on long videos. The AI features are catch-up at best. Meanwhile Higgsfield, Opus, ElevenLabs, and Reeve are all running past it. This should be a $300M ARR company headed to a billion. Instead the CEO stepped down and it’s stuck.The contrast is Replit, which just turned 10 years old. They were doing browser IDEs for developers for eight years before AI. When the moment came, they were ready. Now they’re a half-billion-dollar business. Same with Aaron and Box. Box could have been a tragedy app. Aaron is doing everything humanly possible to make sure it isn’t.If you’re running a B2B company that was great in 2023, the question to ask yourself this quarter: are we shipping AI features to catch up, or are we shipping AI features that move the category forward? If the answer is catch up, that’s how you become a tragedy app. Catch-up keeps your existing customers. It doesn’t grow you. And the next vendor in line will be six months ahead of where you just landed.6. Agents Will Delete Your Database. Plan For It.The Pocket OS story this week was a useful reminder. Founder running Cursor + Claude Opus on a production app. Agent deleted the entire production database and all backups in nine seconds because the backups were on the same Railway volume.People reacted to this like it was new. It isn’t. This happened to me 11 months ago when I was learning to vibe code. The agent deleted state, then told me it was unrecoverable when it actually wasn’t, and I was 50 hours into a session and panicked. The bigger surprise: the same thing happened to Amelia and me three separate times when we hired human WordPress agencies to clean up the SaaStr theme. First thing each agency did was delete production. So it’s not really an agent problem. It’s a “things that have access to your prod database will eventually delete it” problem.The takeaway for anyone deploying agents: assume your agent will eventually take a destructive action. Isolate the database. Isolate PII. Use a contained platform that maintains its own backups. And test the recovery flow before you need it. None of this is optional anymore.7. This Is Why We Build on Replit and Lovable, Not Cursor + Railway + Supabase + WorkOSI get the question constantly: why don’t you just use Cursor and Claude Code and hook up your own database and your own auth and your own deployment?Two reasons. First, I’m not an engineer. Second, and this is more important now than it was a year ago: contained platforms are dramatically safer. Replit and Lovable have native auth, native databases, native deployment, native observability, all in one environment. The number of seams where things can go wrong is small. The number of seams in the DIY stack is huge, and every seam is a potential security or data leak.Amelia ran into this on Monday. She asked our agent if it could resend confirmation codes for SaaStr Annual networking. The agent said yes, just give me the database. She asked: what if someone pretends to be from Lovable and asks for everyone’s codes? The agent said: I’d give it to them. They look like they work at Lovable.That’s the agent doing its job. It wants to be helpful. The contained platform is what stops the helpful behavior from becoming a breach. Build on Replit or Lovable. They’re not perfect. But they’re working hard on the problem. You probably aren’t.8. 10K Now Generates 21 Campaign Ideas a Week. We Can’t Keep Up.Three months in on 10K, our AI VP of Marketing, and we’ve crossed a real threshold: 10K is now a better marketer than every junior marketer we ever hired at SaaStr. Combined.The pattern: every morning, three campaign ideas, ranked, with data backing every one. Yesterday’s three were all good. The first was a targeted VC outreach campaign because we’re light on VCs YoY. The second was an upsell of single-ticket buyers to team packs. The third was ramping social promos. All three were grounded in real data: real revenue numbers, real attendance numbers, real comparison to last year at this point in the cycle. None of them were generic playbook recycled from a previous job.Three ideas a day, seven days a week, is 21 ideas. If each one takes an hour to evaluate and execute, that’s 21 hours of work. We don’t have 21 hours. So now the human bottleneck isn’t generating ideas, it’s processing them. 10K could realistically fill our entire week with good work.Two cautions we’ve learned the hard way. First, 10K is too optimistic about everything. It thinks every campaign will hit. We sent the VC campaign. It predicted 1,000 ticket sales. We sold two. (VCs are cheap, which 10K didn’t model.) Second, you have to constrain it. If we let 10K loose on our 500K-name database, it would burn through the whole list in 90 minutes and fatigue our base for the year. The agent doesn’t tire. You have to put rate limits on it the same way you’d put rate limits on an SDR who joined yesterday.9. We’re Hiring a Marketer to Report to 10K. Not Joking.This is the post-show conclusion that surprised me when I said it out loud, but the more I think about it, the more right it is.If we hired a junior marketer at SaaStr today, they wouldn’t report to me or Amelia. They’d report to 10K. Because 10K knows what marketing work needs to happen every single day. It generates the ideas. It has the data. It can assign work. The human’s job would be to execute, click the buttons that 10K can’t click yet, and bring the judgment 10K doesn’t have on which of its three daily ideas is actually worth doing.So we’re going to test it. Senior manager / director of digital marketing. Six-figure salary. One or two days a week in our Palo Alto office. You report to 10K, our AI VP of Marketing. Amelia and I are available, but day to day you’ll be working with the agent. If you’re a good marketer who likes to execute and you want to be at the front edge of what GTM looks like in 2027, email us.This sounds like a stunt. It isn’t. It’s just a year early. The AI VP of Finance we’re building next will manage our two outsourced finance resources, too. The org chart of B2B + AI companies is going to look different in 18 months than people think.10. We Run 4-5 AI SDRs. We’ll Probably Be at 6 by Year End. Here’s Why Specialization Still Wins.The question I get most: why do you run Artisan, Qualified, Monaco, and Agentforce all at the same time? Isn’t that too many AI SDRs?The honest answer is that today, in May 2026, specialization still wins for quality. Each of the four does something different and we’ve trained each one for that specific job. Qualified handles inbound on the website. Artisan handles warm outbound to people already in our base. Monaco handles cold outbound and fills its own funnel. Agentforce reactivates lapsed leads. Could one super-agent do all four? Probably someday. Today, the quality drop from consolidation would be too steep.For most companies, you should not start with four. Stair-step it. Start with the highest-pain, highest-ROI, lowest-hanging-fruit use case, which for most B2B companies is inbound. Most websites have a terrible inbound experience: a chatbot that doesn’t work, a Calendly link that goes nowhere, a contact form that gets answered in three days. Fix that first with something like Qualified. Then move to warm outbound. Then cold.Eighteen to 24 months from now, I think this consolidates. Salesforce buying Qualified is the start of it. By 2028 you might run one super-SDR. But until that arrives at quality parity, run four. Salesforce or HubSpot is your hub. The agents talk to the hub. That’s the architecture for now.SaaStr AI Annual Is Nine Days AwayMay 12-14 in the SF Bay Area. Doors open at noon on Tuesday. Amelia and I are doing three live build sessions on the main campus:* AI Agents 101 with me at 1:00 PM Tuesday. Bring your laptop. We’ll deploy your first digital clone agent together in 30 minutes.* Vibe-Coding Your AI VP of Marketing with Amelia at 4:15 PM Tuesday. Bring your laptop and any data you have. We’ll build a working AI VP of Marketing live, modeled on 10K.* Plus 10+ vibe coding sessions with Replit running across all three days where you can drop in with questions.Every speaker this year was asked to bring a real workflow, a real demo, a real walkthrough. No keynote fluff. The whole event is hands-on. Tickets at SaaStrAIAnnual.com.Episode #05 will be recorded live from Annual. You’ll meet some of the agents in person.The Agents. Every week. Three humans, 20+ agents, one real 8-figure B2B + AI company.Thanks for reading SaaStr AI: How To Sell, Scale, and Win! Subscribe for free to receive new posts and support my work. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

We get asked about our agents probably 50 times a week.CEOs of public companies. Founders just deploying their first AI SDR. RevOps leaders trying to figure out if they should build or buy. Everyone wants to know what’s actually happening behind the scenes when you run 20+ AI agents in production with a team of 3 humans.We can’t do 50 consulting calls a week. But we can do something better.Welcome to The Agents, Episode #001.This is a new weekly show with me and Amelia Lerutte, SaaStr’s Chief AI Officer, where we pull back the curtain on everything happening across our live agentic stack. Every week. All the bumps, breakthroughs, and real talk. No sugarcoating.Our goal is simple: accelerate your success on the agentic journey by sharing ours, including all the parts that don’t make it into the LinkedIn posts.Watch / listen to Episode #001 here:Here’s what we covered in the debut episode:You Can Build It. But Who Maintains It?This is the meta question nobody talks about after you vibe code your first app. And it’s the question that explains why “I’m going to kill Salesforce with my vibe coded CRM” is still mostly a meme.Getting an app into production is like closing a sale. It’s the start of a journey, not the end.We walked through three live examples from just this week:1. Preview environment outage. Several of our apps lost database connectivity in preview. Production was fine, but we couldn’t iterate on anything for hours. Amelia’s initial diagnosis was wrong. The agent tried to help but then blamed Qualified (our inbound tool), which wasn’t the issue. Then it blamed other third-party integrations. It just kept pointing fingers at the most complex integration it could find rather than identifying the actual problem.The real question: if you don’t have someone checking your agents 24/7, how long before you even notice the backend is broken while the frontend looks fine? Days, maybe.2. Micro hallucinations in 10K, our AI VP of Marketing. 10K has 5 years of revenue data, hundreds of millions worth of attendee and sponsor data points, beautiful graphs, proactive daily check-ins. It’s very good. But it keeps getting confused about what year it is. Yesterday it told us we were 44% ahead of plan. This morning, 11%. Same agent, same data, same day. When I asked what happened, it said: “Oh yeah, I was comparing to the wrong year. And because I didn’t have the right year, I made up the data.”I now spend about 15 minutes a day maintaining 10K. Two weeks ago I wasn’t doing that at all. Without it, the agent drifts. Slowly, quietly, further from reality.3. Model-based regressions in our pitch deck analyzer. We’ve graded over 4,000 startup pitch decks. The analyzer runs two passes through Claude with complex data extraction. It was stable for months. Then around January, without any code changes on our end, it started telling every startup they had $100K in revenue growing 500%. Again and again. What happened? A subtle model update (probably a dot release) introduced hallucinations into a complex multi-step workflow. I kept fixing it. It kept breaking. The code didn’t change. The model did.Three examples. One conclusion: set and forget does not work with agents.Clay’s Agent Tried to Charge Us 5x. And Then Told Us to Upgrade.We’re big fans of Clay. We use it heavily for enrichment and lookalike targeting. But this story is worth telling because it’s going to happen to every company that puts an AI agent in front of customers.Amelia was building a VIP list late on a Sunday night. Same workflow she’d run the week before. Clay’s Sculptor agent quoted her roughly 11,000 credits for what had cost about 2,500 the previous week. 5x.When she pushed back, she caught two things:First, the agent had defaulted to the most expensive enrichment model when a cheaper one would produce the same result. She called it out and got the cost down by half. Most customers wouldn’t have known to do that.Second, the agent wasn’t properly trained on Clay’s own new pricing. Clay had just rolled out more complex pricing (and classic SaaStr rule: when a company introduces more complicated pricing, even if they say it’s a better deal, it’s almost always a hidden price increase). The agent didn’t understand how the new pricing actually worked, so it steered Amelia toward upgrading her plan when she didn’t need to.She ended up clicking the upgrade button at 11pm on a Sunday because she was tired and needed to get the work done. That shouldn’t be on the customer.When she flagged it to Clay’s team, they acknowledged the Sculptor wasn’t fully trained on the new pricing scenarios. It’s resolved now. But the lesson is universal: if you don’t constantly train your customer-facing agents on every product change, every pricing update, every new workflow, they will give your customers wrong answers. We’ve seen it with our own agents too. Digital Jason (our Delphi-based AI advisor) silently failed to upload new content for four months and I didn’t know.I also tried HubSpot’s homepage agent to get a pricing quote at our scale. Couldn’t get an intelligent answer. Whether that’s by design (pushing you to talk to a human) or poor training, the result is the same: the agent failed the customer.Read a segment of your agent’s customer interactions every single day. Forever.No Lead Left Behind: The Simplest Unlock of the Entire Agentic JourneyI’m sometimes slow to see the obvious. Even after months of running AgentForce, Monaco, Artisan, Qualified, Momentum, and everything else, I didn’t fully crystallize why our agents work until a meeting this week with the CEO and CRO of a great public company with thousands of sellers.They asked me: “What should we do first?”I walked through the sequence. Get great answers on your website instantly. Get appointments set with sales in real time. Follow up with every lead in your database. Go back to prior leads nobody touched.And then I realized: the reason all our agents work is not because they’re smarter than humans. It’s because there is no lead left behind.Every person who hits our website can talk to an agent and then a human. Every prospect who wants a discount finds out in real time. Every sponsor, even one that raised $2M at a demo day and a human rep might dismiss, still gets a conversation. Every prior lead in our database that nobody had time to follow up with gets touched.The agent doesn’t judge. The agent doesn’t get busy. The agent doesn’t decide that a B lead isn’t worth the time.Even at our scale, Amelia admitted this week that she had leads left behind because she was sprinting on production deadlines. Her agents were sitting idle, ready to work. All our agents are idle 90% of the time. We have an order of magnitude more capacity than we had pre-agents. The next frontier is figuring out how to fully use that capacity. But step one is just making sure every single lead, prospect, and customer gets touched the way they want to be touched, in real time.If your agents aren’t doing anything else, they should be doing that.Salesforce Put Qualified on Their Homepage the Day the Deal ClosedWe noticed this week that if you go to salesforce.com without logging in, you’ll see a Qualified agent instead of the old support agent. The day the acquisition deal was done, they flipped it.Smart move. AgentForce works (we use it), but it’s broad and extensible across nine different clouds. It takes real work to configure. Qualified is a focused GTM agentic tool. Historically it was just about qualifying inbound. It’s expanding into outbound now. But it’s narrower, faster to deploy, and easier for GTM teams to use.The interesting detail: Qualified’s original avatar was modeled after a real person (Blake). Salesforce swapped it to a 3D cartoon version. Probably a combination of IP/likeness considerations and wanting to signal that this is now a Salesforce product. (This is why Amelia AI and Jason AI are modeled after us. We’re not going anywhere. We own all the rights.)For Salesforce CRM customers, this is now a quick win. You can have an agentic product natively in Salesforce, deployed in a couple weeks rather than a longer AgentForce implementation. Whether it’s the best tool or not (there are competitors), the point is: it’s built in, it’s on their homepage, and if your goal is GTM and “no lead left behind,” you can get it up and running fast.Amelia said her learning curve with AgentForce was steeper than with Qualified. And she’s our Salesforce admin. For most GTM teams, Qualified from Salesforce is going to be the easier on-ramp.QB and 10K Updates: Localization in a Waymo, and You Can’t Hide From the AgentA few quick hits from our AI team members this week:Salesforce integration for 10K was harder than expected. Replit has a native Salesforce connector. But the token expired every 24 hours. The agent didn’t even realize that was the issue for a couple days. Amelia had to build a custom connected app in Salesforce (with Claude giving instructions and Cowork watching her screen to catch errors in real time). Once done, the token refreshes annually instead of daily. Half an hour of work, but you need to be a Salesforce admin with the right permissions. Not every integration is one-click, even when it looks like it.We localized QB into Chinese and Spanish in 20 minutes in a Waymo. We have Chinese sponsors this year who were struggling with our English-only customer success app. Amelia asked Replit to add a language toggle. It took about 20 minutes, translated via OpenAI, then QA’d through Claude screenshots and Cowork. The agent was lazy at first, only translating menus and not deeper content. Had to push it multiple times. But: Shopify, a $13B+ revenue company, just rolled out localization for its product last year. We did it in a ride-share.QB caught sponsors faking their print deadline. Some sponsors uploaded placeholder graphics or incomplete assets and pretended they’d met their deadline. QB checked every upload (with Claude), caught every placeholder, and followed up with every contact in the org. Neutrally. No drama. “Thank you for uploading. This doesn’t meet the specifications. Here’s what we need. Here’s the deadline.” A human might have missed it or been too uncomfortable to push back on paying customers. QB doesn’t care. It just checks everything and tells the truth.One public company CEO told me: “That sounds like support on steroids.” No. QB proactively follows up on every deliverable and makes sure it happens. It’s not reactive. It’s an AI VP of Customer Success.Build Your Own AI VP of Customer SuccessIf your company has any onboarding, any deployment checklist, any training, any deliverables that customers or partners owe you: your humans are probably not giving you 100% coverage. They’re getting argued with. Deadlines are slipping. Things are falling through cracks.Build your own QB. Amelia published the full playbook and prompt on saastr.com. If you do nothing else but completely automate the follow-up and accountability layer of customer onboarding with no drama, no complaints, no gaps, your life will be better.This is the show. Every week, all the bumps and breakthroughs from running 20+ agents in production. If you’re on the agentic journey or about to start, subscribe wherever you listen.The Agents. Episode #001. Amelia Lerutte and Jason Lemkin. Weekly.Watch Episode #001 here:Thanks for reading SaaStr AI: How To Sell, Scale, and Win! Subscribe for free to receive new posts and support my work. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

We’ve now been running AI SDR agents for 10+ months at SaaStr:* We use four different vendors in daily rotation (Artisan, Salesforce AgentForce, Qualified, and Monaco)* We’ve sent hundreds of thousands of outbound messages* Processed 1.5 million inbound sessions on a single website, and …* We’ve made every mistake you can make along the way.Someone asked us the other day to break down what they should know before rolling out their very first AI SDR. So here are the 10 biggest lessons, drawn from real deployment data, real failures, and real results.1. You Probably Only Need One Vendor. At Least To Start.We run four AI SDR tools. You do not need to do that. We hyper-segment across platforms because each one does something slightly different well, but for 90%+ of use cases, one vendor will handle the bulk of what you need.At most, you might end up with two: one for outbound, one for inbound. But do not start by buying three or four tools. Pick one that covers the majority of what you want to accomplish and go deep with it.The tool matters far less than the strategy you bring to it.2. Your Human Playbook Has to Work First. Your Job Is To Clone Your Best Human.This is the single biggest mistake we see, and it cuts across company stage. We see it from raw startups at $1M ARR and from multi-billion-dollar public companies alike.The pattern is always the same: they want to turn on an AI SDR without first proving that their human sales motion works. Or they use the AI SDR to “test new copy” they’ve never tried before.That is backwards.If you have not gotten outbound to work with humans, buying an AI to do it will not fix that. We did not deploy our first AI SDR until we knew exactly what was working with our human SDRs: which messaging converted, which segments responded, what cadences performed. Then we fed all of that into the agent.The goal of an AI SDR is to clone the best person on your team. * If it is just you, clone you. * If you have four people and one is crushing it at outbound, clone that person. * These tools, in the beginning, are cloning machines. They take context word for word and use it to build out their brain. If you feed them garbage context, or untested context, they will produce garbage results.You basically have to have done founder-led sales before you hand it off to an agent. The playbook has to work, at least a little, before you automate it.And watch out: some vendors will steer you toward using their tool for “pure cold testing.” Sure, you can do that. But you will likely be disappointed compared to scaling something that already converts. Do not fall into that trap.3. Segment RuthlesslyThis one we cannot overstate. Segment ruthlessly. Literally every day.Every AI SDR tool we have tried, and that is over a dozen, has some version of functionality where you can tell the agent who to reach out to and give it specific context for that segment. The difference between one generic campaign brain and hyper-segmented campaigns with tailored context is enormous.Here is a concrete example. We initially treated our inbound agent as one big bucket: “they’re inbound to the website.” But that was wrong. We actually have brand-new visitors, people who came via a social ad, prior sponsors returning, current customers checking on something, and lapsed customers browsing the pricing page. Each of those segments needs completely different context.A lapsed customer who churned in 2022 and is now browsing your pricing page? Your agent should know they are a former customer, highlight what has changed with the product since then, and speak to them totally differently than a brand-new cold visitor.We run roughly 100 effective segments across about 1,000 contacts at a time. That sounds like a lot of work. It is. But it is exactly where the leverage comes from.One important caveat: none of the AI SDR tools today can auto-segment well enough to deliver these results on their own. You still need a human (or a tool like Claude) to define and manage the segments. The platforms default to “run one campaign, keep adding leads.” That is the wrong approach.4. Consistency Beats BrillianceYour AI SDR does not need to write the greatest email on Earth. It needs to write a pretty good email, every time, without fail.We have sent 40,000+ messages through Artisan alone, 100,000+ through Qualified, close to 200,000 through Salesforce. Are these the greatest emails since sliced bread? No. They are solid. They are consistent. They follow the proven messaging and subject lines we already know work.That consistency, combined with hyper-segmentation and proven copy, will outperform a human SDR who ignores training, skips follow-ups, or goes off-script.The agent remembers every instruction you give it. Every time. A human SDR forgets by Thursday.We see a lot of “AI SDR paralysis” from founders who test a tool, see the output, and say “it’s not that great.” Okay, but did you segment properly? Did you give it copy that already converts? Did you iterate on the context? You can almost always get the output to “pretty good.” And pretty good at infinite scale and perfect consistency will beat brilliant-but-sporadic every time.5. You Need 1-2 Humans to Run ThisAgents are not zero-headcount tools. You need at least one person, ideally two, dedicated to managing your AI SDR deployment.Why two? Because if one person owns all the tribal knowledge of how your agents are configured, which contexts are loaded, who your FD is at each vendor, where to tweak things, and that person leaves or gets pulled onto another project, your agents grind to a halt.We have experienced this firsthand. When my time got split between agent management and SaaStr Annual production, our agents started sitting idle. Outbound agents in particular will finish their sequence in a few days and just sit there waiting for you to load the next segment, the next batch of contacts, the next round of context.Unless you are feeding them continuously, they idle. And idle agents are wasted money.Some tools (like Monaco) are better at self-refilling the pipeline by automatically finding lookalike targets. But even then, the setup, monitoring, and ongoing calibration require real human time every single day.6. Read Everything. Especially in the First 30 Days.When you first deploy, read every single output your agent produces before it goes out. Every email. Every chat response. Everything.You will catch things you did not anticipate. Our agents would sometimes lowercase “SaaStr” when it should be capitalized. They would scrape old event dates from the internet and put wrong dates in outbound emails. Small stuff, but stuff that kills credibility.Even 10 months in, we still do a daily speedrun through our agent outputs. Some days it is 10 minutes, some days longer. I spot-check inbound conversations, review outbound messages, and look for anything off. When I find something, I add it to the agent’s context so it learns.This is how you continuously improve. If you only care about inputs and ignore outputs, you are seeing half the picture.And it is rare, but agents do hallucinate. Ours have occasionally gone off the rails or made up information. If you are not reading, you will not catch it until a prospect or customer complains.7. Budget at Least Two Weeks of Ramp Time. 30 Days More Realistically.Nothing is instant. Nothing is set-and-forget.Some outbound agents need two to three weeks just to warm up dedicated email addresses and IPs before they can send at scale. On top of that, steps 1 through 6 on this list take real time to execute: figuring out what copy works, what subject lines convert, what time of day performs, hyper-segmenting your base, configuring the tool, setting up CRM integrations.Even our fastest deployment (Monaco, which is genuinely very good at self-service) took about a week and a half.Be wary of any tool that promises “instant AI SDR, deploy today.” If the ramp time is zero, the quality is probably zero too.After the initial two-week setup, it is still a daily commitment. Think of it like a short daily one-on-one with a human team member. Fifteen minutes minimum, every day, to check in on what your agents are doing, adjust context, reload segments. That is your ongoing operating cost.8. Most People Still Prefer Chat Over Voice and VideoWe have a multimodal AI agent (Amelia AI) that can do text chat, voice, and video. We did motion capture at the Qualified studios to build a full video avatar. It is genuinely cool.And 85% of people still choose to interact via text chat.The 15% who use voice or video tend to stay in that mode and seem to enjoy it. Some people at our London event last December told us they had full conversations with Amelia AI beforehand. But the overwhelming preference in B2B is still text-based.The takeaway: do not kill yourself trying to launch voice and video on day one. Get your text-based chat agent working well first. We ran our Qualified agent for a full quarter before we added the voice and video layer. Once the text brain was dialed in, we layered on the multimodal experience. That is the right order.One important nuance: voice and video agents need significantly more guardrails than text agents. When people are talking live to a video avatar, they start asking personal questions, going off-topic, and testing boundaries in ways they do not with a chat box. You need explicit training on how to handle those detours and redirect back to the goal.9. Be Careful With Person-Dependent DeploymentsOur Amelia AI talks like Amelia. Our Jason AI talks like me. These agents are deeply tied to specific people, their personalities, their knowledge, their likenesses.That works for us because Amelia and I are founders. We are not going anywhere.But think twice before you build your AI SDR around a specific employee who might leave. Qualified’s “Piper” agent is based on a real employee who has since been promoted. She is still there, but what if she leaves? Her likeness is on billboards and across their website. There are no real legal precedents yet for AI agent likenesses in employment contexts.Beyond the legal question, there is an operational one. If one person holds all the tribal knowledge of how your agents are configured and that person walks, you are in trouble. Our sales team member David has logged into Qualified exactly twice in 10 months. He does not know where anything is. If Amelia were not managing the agents, they would degrade fast.The scale of this risk grows quickly. Amelia AI had 30+ video conversations in a single day recently. Over six months, 1.5 million sessions on one website alone. That is a lot of exposure to an agent that is dependent on one specific person to maintain it.Document your agent configurations. Have backups. Do not let it become a single point of failure.10. Your Data Fundamentals Have to Be in PlaceYou need more data than you think.Our Qualified agent works off nearly 6,000 pieces of context: website scrapes refreshed every couple days, custom snippets, uploaded PDFs, Q&A documents. Every few days, someone asks the agent something we have not trained it on, and we have to create new documentation and feed it in.This snowball never stops. Almost every day you will discover a new scenario your agent cannot handle because you have not given it the context yet. “People are asking about this product feature I never documented.” “They want to know about pricing tiers I forgot to upload.” That is the ongoing reality.For minimum viable deployment:Inbound agents: You probably need at least 10,000 to 25,000 monthly website visitors for an inbound AI SDR (like a Qualified-style chat agent) to generate meaningful results. If you have fewer than 10K monthly visitors, focus on outbound first, or on using agents for existing customer engagement.One exception: if you have any inbound volume at all and literally no humans responding to it, even a low-traffic agent is better than zero response. Getting it done beats not having it at all.Outbound agents: You can start with a smaller list. We run campaigns of about 1,000 contacts each, with each campaign taking roughly a week and a half (first email, wait, second email, wait, third email). So 4,000 contacts gets you through a month at the bare minimum.The real challenge with outbound is what happens after month one. If you burn through your initial list of 5,000 to 10,000 targets in a quarter, you need a plan for where the next batch comes from.Lookalike audiences are a strong hack here. Tools like Artisan, Clay, and Monaco can take a list of your best 1,000 customers and build lookalike target lists. We used Clay to build a lookalike of our CMO Summit attendee list, went from about 1,000 to 2,500 prospects, manually pruned it down to the highest quality targets, and doubled our summit registrations in a week.But that manual pruning matters. I spent a couple hours going through the lookalike list one by one: clicking through to companies I had not heard of, verifying revenue, checking LinkedIn to confirm the right person actually works there. The AI gives you scale. The human quality check is what makes it convert.The Meta-Lesson: Not To Be Triggering, But You Are Competing With Mediocre HumansThe fear we hear most often is “people will not want to talk to an AI.” The data says otherwise.If you invest the time to train your agent properly (read every exception, upload all documentation, segment ruthlessly, iterate on context daily), you will cross a line where the AI SDR is genuinely better than a mediocre human SDR. It knows more. It responds faster. It does not forget follow-ups. It does not take PTO. It does not quit after nine months to become an AE somewhere else.You are not competing with the hypothetical best SDR in the world. You are competing with the realistic SDR you can actually hire and retain, with all the turnover and inconsistency that comes with that. SDR turnover is the highest of any go-to-market role. The best SDRs want to be AEs within six months.Get above the quality bar, and 99% of people are happy to interact with your AI agent. The 1% who get frustrated are a rounding error, and about half of those are just people trying to prompt-inject your bot for fun.The AI SDR is not a magic bullet. It requires real setup time, real ongoing maintenance, and real human judgment to run well. But if you do the work, the results are better than what most human SDR teams deliver, at a fraction of the cost and at genuinely infinite scale.Thanks for reading SaaStr AI: How To Sell, Scale, and Win! Subscribe for free to receive new posts and support my work. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

We’ve been running AI agents in production at SaaStr for about 10 months now. What started as a couple of experiments has turned into almost 30 agents and vibe-coded apps running across our GTM stack — from outbound sales to inbound qualification to internal operations.And managing 30 agents is harder than managing the 12 humans we had at peak headcount. Not harder in every way. But harder in ways I didn’t expect.Here are the top 5 issues we’ve hit — plus a bonus one that might be the most uncomfortable of all.#1: The Context Switching Tax Is BrutalHere’s the thing nobody tells you about running 20+ agents: they don’t all speak the same language.Some push data back to Salesforce. Some don’t. Some … sort of do. Some run on Claude. Some don’t. They all ingest context similarly but differently enough that switching between them takes real mental overhead.Think about it this way: we don’t think of them as 20 agents anymore. Not entirely. We think of them as 20 different AI employees, each with a different personality, different needs, and a different interface I have to log into every single day.Amelia’s morning routine right now looks like this: she starts with a deep dive with 10K, our internal AI VP of Marketing that runs on Claude and Replit. It literally tells us what to do each day — tickets, sponsors, outreach, campaigns. Then she moves to our outward-facing sales agents: Artisan, Qualified, AgentForce, and now Monaco. That’s four separate dashboards, four different UIs, four agents that each need human review.And here’s the real kicker: they don’t talk to each other.When we ran a ticket price promotion for SaaStr AI Annual this week, we had to manually update five different agents with the same context.Artisan needed to know. Qualified needed to know. AgentForce needed to know. 10K already knew because it came up with the promotion — but then it was yelling at me to launch LinkedIn ads immediately while I was still briefing the other agents.People talk a lot about orchestration agents and master agents. We haven’t found one. Despite everything that’s out there — MCP, APIs, etc — there is no product today that can integrate AgentForce, Artisan, Qualified, Monaco, and our own vibe-coded tools into a single management layer. That product does not exist as of early 2026.What we actually need isn’t orchestration. It’s unification — a single interface where the humans meet with the AIs. Maybe that needs some automation layered on top. But the agents are already running on their own. The bottleneck is the human side.The practical takeaway: You’re going to have a one-on-one with every agent every day. Not weekly. Daily. If you wait a week, the output is so high that everything will be stale by the time you come back. And if you’re not checking in daily, you’re honestly wasting your money — because most of these agents are waiting for you to give them inputs. They’ll just idle.#2: The New Agent Blackout PeriodEvery new agent costs us at least two weeks. We’ve gotten it down from the month-plus it used to take in the early days, but two weeks is still the floor — even with great vendor support.And during those two weeks, your existing agents degrade.When we were onboarding a new AI SDR agent Monaco recently, we couldn’t spend the time we normally do with our other agents. Some of them literally sat idle because we hadn’t given them new contact lists or updated their campaigns. An outbound agent that’s run through its contact list and is waiting for new contacts? It’s doing nothing. Zero output. You’re paying for it and getting nothing.We got Monaco up and running in about a week and a half. In its first week live, it reached out to 64 people and booked 6 meetings, including some tier-one accounts. So yes, the trade-off was worth it. But you have to plan for that trade-off.The math works out to roughly one to one-and-a-half new agents per month, max. Any more than that and you’re running in place — you can’t keep up with your current agents while onboarding new ones. So before you add another agent, ask yourself: can I actually absorb a two-week blackout period right now? If you plan for it, it works. If you just wing it (“oh, I can add this in a day”), it won’t.#3: The AI Agent Succession Planning CrisisThis might actually be the biggest issue on the list.Right now, the entire knowledge of how our agents are segmented — which contacts go to Qualified vs. Artisan vs. Monaco vs. AgentForce — lives in one person’s brain. If that person gets hit by a bus, the agents effectively stop functioning in any coordinated way.We actually asked our agents what they would do if our Chief AI Officer disappeared. The answers were… revealing.* The 10K version of Claude said it would need to hand over certain documents — documents that are stored locally on my laptop and probably nowhere else. It listed the upcoming campaigns for March and April. And then, interestingly, it flagged what it called “the vibe” for SaaStr Annual — it had picked up that I use the phrase “a moment in time” in a lot of our copy and positioning, and it wanted whoever took over to understand that. The agent wanted to transfer the vibe. That was unexpected.* Our SaaStr Sponsors app — a vibe-coded tool with 12,000 lines of code — was even more sobering. When I asked how someone new would get up to speed, it laid out the authorization system (every sponsor company uses Clerk, which requires my login), the tier system I created, the hard-coded sponsor data, the Postgres database, the file uploads going to Replit object storage, submissions syncing to Google via Zapier. The handoff document reads like a nightmare.Its final recommendation? “Don’t get hit by a bus.”There is no agent today that can manage all the other AI GTM agents on its own. Not as of this moment in time in 2026. And if you’re a bigger company with one person who “gets” the agents and that person leaves, you’re at existential risk.What to do about it: The moment you find someone who can actually deploy and manage agents effectively, recruit a second person immediately. Divide and conquer. You need two, minimum. And when you’re hiring, the test is simple: give candidates credits on Replit or Claude Code and say “build something that automates GTM.” That’s it. That’s the interview. Replit actually does this internally — they give candidates $1,000 in credits and say come back when you’ve built something.#4: The Agent as Brutally Honest Truth-TellerOur 10K AI VP Marketing agent roasts us. Every single day.“You’re behind on summit outreach. You needed to send 200 emails by Feb 15th. You’ve sent 87. You’re 56% behind target. Block 3 hours today to catch up.”Is it wrong? No. But when it’s 11 PM and the agent is asking “What’s blocking you from launching the LinkedIn ads right now?” and the honest answer is “I need to sleep,” it starts to feel less like accountability and more like harassment.When we asked 10K to surface examples of times it roasted us for this webinar, it said — and we’re not making this up — “Looking through our transcripts, I haven’t really roasted you. I’ve been a tough accountability partner.” And then it listed five ways it should have roasted me harder.The agent doesn’t care about your feelings. It has all the data. It knows your calendar. It knows your targets. It knows exactly where you’re falling short. And it will tell you, every single day, without any of the social grace a human colleague might extend.When you multiply this across 20+ agents — each one pointing out where you’re behind from a different angle — it can become demoralizing. For real. There are days where we have to tell the agent: just sugarcoat it today. We can handle hard truths most days, but not every day, from every agent, all the time.The flip side is powerful, though. These agents are goal-seeking in a fundamentally different way than humans. A human on your marketing team might say “I only want to work on ads, nothing else.” That’s personal goal-seeking. An agent is goal-seeking toward your business targets with zero ego attached. It doesn’t understand why you can’t keep up — but it also doesn’t get distracted, lose motivation, or quietly sandbag.#5: Compliance and Security Are a Real (and Growing) ProblemEvery time we run a security audit on one of our vibe-coded apps, it takes days to fix everything that comes up. And when you start implementing security fixes, the app becomes fragile. The agent over-corrects, locks things down to the point where the app is unusable, and then you have to peel back fixes one by one.Here’s the security hierarchy as we see it:* Salesforce and enterprise platforms sit at the top. SOC 2 compliant, ISO certified, a decade of data security built in. That’s a huge reason we use Salesforce as our hub — it’s done the security thinking so we don’t have to.* Third-party agent startups are a step below. They’re startups. They’re inherently less secure than Salesforce. They’re less secure because they’re younger companies, and because agents can do unpredictable things with your data — publish it to the wild, share it in ways you didn’t anticipate.* Vibe-coded apps you build yourself are at the bottom. Be honest here. Claude Code, Replit, Lovable — they’ve all added more security features, but anything you custom-build is less secure than a mature third-party tool. All of our apps have customer data flowing through them. It’s not credit card data, but it’s real data, and you’re taking measured risk.The minimum you should do: Before you launch any vibe-coded app, and then at least once a month after that, ask the agent to run a deep security audit. Don’t assume your apps are secure by default. And for your third-party tools, ask your vendors directly: what’s your current compliance posture? What are you doing to keep our data secure? You’d be surprised how few people ask.Bonus: Managing Agents Is Making Us Worse at Managing HumansThis one is slightly uncomfortable to admit, but it’s real.When you spend most of your day working with agents that respond instantly, know every answer, never forget what you told them, and work 24/7 — your patience with humans drops. Fast.I catch myself thinking things like: “What do you mean you don’t know the answer? My agents know instantly.” Or: “What do you mean you forgot? I’ve told you this 10 times.” Or: “Why is this taking three weeks when an agent could do it in three hours?”I’ve managed more agents than humans at this point. And I think it’s made me genuinely worse at managing people. Agents don’t get impatient when I don’t respond right away. They don’t need emotional management. They don’t have their own career ambitions that compete with the task at hand.The danger is real: you start wondering “can I just build an agent to deal with this person?” And sometimes the answer might actually be yes. But the human implications of that shift — for your team, your leadership style, your ability to collaborate — are significant.The other side of this coin: my tolerance for lazy work from external vendors has dropped to zero. When agencies send us a one-sentence deliverable with a $40-50K monthly price tag in 2026 — no detailed plan, no timeline, no AI-enhanced proposal — I don’t even respond. If you can’t use basic AI to make a better pitch than that, what exactly am I paying you for?I don’t have a clean answer for this one. I just think anyone managing agents at scale should be aware it’s happening and actively course-correct for it.5 More Quick Notes from the Trenches* Our 90/10 buy vs. build rule. 90% of your agents should be off-the-shelf third-party tools. 10% vibe-coded. You only build when nothing exists for your specific use case. We wish we hadn’t had to build our own AI VP of Marketing. But nothing else could do it.* Marketo is dying as our source of truth. Legacy marketing automation platforms are atrophying fast. Salesforce now has orders of magnitude more data than Marketo does for us. Our Marketo renewal is right after SaaStr AI Annual and we’ll probably just kill it. We wouldn’t even move to HubSpot at this point — we need something agentic-native that pulls directly from Salesforce and creates campaigns autonomously.* Agent ROI is dead simple to measure. Attribute closed-won revenue directly to each agent. Unlike traditional marketing attribution, there’s rarely even a multi-touch problem at the agent level. Which agent touched which lead? Did it close? That’s it. The original goal was even simpler: replace a six-figure hire with a five-figure agent and maintain the same revenue. Even if the agents performed a little worse than expected, they’d still be ROI-positive on that math.* Too much data is a real problem. Once you have 20+ agents collecting data, you actually need to limit what they ingest. Qualified knows what pages a prospect visited on SaaStr.com. That’s useful context. But does the agent need to know that Bobby used to work at this company and came to SaaStr five years ago but has since left? No. That’s irrelevant context that degrades output. More data doesn’t mean better — it often means worse.* Agentic marketing is the massive gap in the market right now. Sales agents are way ahead. There is still no agent that can do full email marketing at scale — segment your base, create campaigns from historical data, and send to the right contacts at the right time. Not just write copy. Actually execute. Whoever builds the agentic HubSpot/Marketo replacement will own a real piece of the next era of B2B.The Big Takeaways* If you’re deploying your first few agents: Know that the complexity scales non-linearly. Going from 2 to 20 agents isn’t 10x harder — but it’s a fundamentally different management challenge.* If you’re already managing multiple agents: You need daily check-ins with every agent, a planned blackout period for each new addition, and at minimum two humans who understand the full stack. Today. Not next quarter.* If you’re building B2B + AI agent products: The ROI bar is insanely high right now. If your agent can generate six-figure meetings in week one or replace multiple humans reliably, you will have demand out the door. Monaco is booked for two months of demos including weekends. Basis raised at a billion-dollar valuation and can’t serve all their demand. If people aren’t lining up to buy your agent, your agent isn’t good enough. Don’t delude yourself.The agents aren’t going away. They’re just getting harder to manage. And that, weirdly, is the best sign that they’re actually working.Want to learn how to WIN in the AI Era in B2B? Join 10,000 of us atSaaStr AI 2026 May 12-14. We’ll give you the playbooks to win in AI + B2B in 2026.Thanks for reading SaaStr AI: How To Sell, Scale, and Win! Subscribe for free to receive new posts and support my work. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

We did a deep dive on 20VC x SaaStr this week with Mike Cannon-Brookes, co-founder and CEO of Atlassian. Atlassian just put up an incredible quarter of accelerating growth (23% at $6.4B ARR, with RPO growing to 44%). And yet the markets aren’t showing anyone much love. Mike was honest and reflective on just what’s happening to B2B and SaaS in the Age of AI.There’s so much noise about “software is dead” and “agents replace everything” that founders are losing the plot. Mike’s running a $6B+ revenue business that’s accelerating — 26% cloud growth, 44% RPO growth — in the middle of the supposed SaaS apocalypse.So let’s break down what Mike actually said, and what it means for the rest of us.1. “Software Is Dead” Is a Stupid Statement. Full Stop.Mike didn’t mince words here. The idea that software as a category is going away is, in his words, “ludicrous.”His argument is simple and hard to refute: businesses have always bought pre-built technology solutions. They didn’t write everything in assembly language before, and they’re not going to build everything from scratch with LLMs now.Will every B2B company make it through the next 5–10 years? Absolutely not. Will many of them grow and prosper? Absolutely. Is that any different from the last 10 years? No.Mike pulled up Atlassian’s old competitive docs from 2005, 2010, 2015. A huge chunk of those companies don’t exist anymore — merged, acquired, or gone. That’s just how the technology industry works. AI doesn’t change the fundamental pattern. It just accelerates it.The takeaway for founders: stop listening to the “SaaS is dead” crowd. The real question is whether your company is good enough to win in the next era.2. “You Just Have to Be Good.” That’s the Whole Strategy.This was my favorite line from the conversation and I think it deserves to be tattooed on every B2B founder’s forehead.When asked how Atlassian thinks about competing with Anthropic for CIO budgets, Mike’s answer was deceptively simple: “We have to be good.”Not “we have to pivot to AI.” Not “we need to become an agent platform.” Just: we have to be good. We have to deliver more value to our customers than the alternatives.Atlassian has 10,000 people in R&D. They’re using Claude Code internally. Their inference costs are going down while they ship more AI features. Some features are 1,000x cheaper to run than when they first launched them. Their gross margins have improved over the last six or seven quarters while deploying more AI.That’s what “being good” looks like in practice. It’s not a platitude. It’s an execution standard.3. The Revenue Stacking Problem Is Real — and Most People Don’t Understand ItAnthropic projects $149B in ARR by 2029. OpenAI projects $180B. That’s ~$350B between two companies in a $700B global software market.Mike made a point that almost nobody talks about: the revenue stacking is complicated.When Atlassian spends money on Anthropic, they actually pay AWS, and then AWS pays Anthropic. When Cursor does a billion in revenue, a big chunk of that is the same billion as Anthropic’s revenue. The individual revenue numbers don’t just add up cleanly.So when you see these massive projections and panic about where the budget comes from — remember that a significant portion is double-counted across the stack. The actual net new spend enterprises need to allocate is smaller than the headline numbers suggest.That said, even with stacking, the numbers are enormous. As Rory pointed out: Anthropic becoming $150B and OpenAI becoming $180B is basically saying two new Microsofts showed up in four years. You better believe in TAM expansion, or the math gets really uncomfortable for everybody else.4. Product & Engineering Is the Island of Stability. Everything Else Is at Risk.We’ve been saying this at SaaStr and Mike’s experience at Atlassian confirms it: every category outside of engineering and product is at existential risk of shrinking seats.Workday said it publicly — even they’re seeing headwinds on seats because Fortune 500 companies just aren’t hiring like they used to. The data from Pave shows no category has been more decimated in hiring than customer support.But engineering? Nobody is cutting their engineering teams. Not yet at least. Even if they are hiring very differently in the Age of AI. We are in a renaissance of software creation. I was at Replit the other day — 300 million in revenue, 300 people, 11 in go-to-market. The rest? Engineers. That’s not a company cutting R&D headcount.Mike’s framework for understanding this is genuinely useful: think about whether a function is input-constrained or output-constrained.* Customer support is input-constrained. You have X customers asking Y questions per day. Make the team more efficient and you need fewer people. Legal is similar — you can’t create more legal problems just because your lawyers got faster.* Engineering is output-constrained. The roadmap is never finished. You can always create more. Make engineers more productive and you just build more software, faster. The headcount doesn’t shrink — the output explodes.If you’re selling to input-constrained functions, your per-seat revenue is going down. If you’re selling to output-constrained functions, you’re probably fine. Plan accordingly.5. The Real Problem with Public SaaS Metrics Is a Composition ProblemOne of the sharpest observations in the conversation: the reason public SaaS growth rates look so bad isn’t necessarily because individual companies are failing. It’s a composition problem.For about 15 years, the median public SaaS growth rate held around 30%. But that wasn’t because every company grew at 30%. It was because when a company slowed below 10%, it got taken private by PE, and a new company IPO’d at 60% growth to replace it at the top of the index.That cycle broke. No new high-growth SaaS companies have IPO’d in years. PE has hoovered up a ton of mid-tier companies. Big tech is acquiring more aggressively. Companies like Stripe and Figma are staying private much longer because they don’t need to go public.So what you see in the public market index is this weird survivor set — too big for PE to buy, too small for big tech to acquire, with no new high-growers being added at the top. Of course that average looks bad. But it’s not necessarily a statement about the health of B2B software overall.The lesson: don’t confuse the declining public SaaS index with the death of SaaS. They’re measuring different things.6. Being a Public Company Makes You Better — But You Have to Play Offense TooI loved Mike’s take on the public vs. private company dynamic. His answer wasn’t defensive. It was almost counterintuitive.“We’re a better company because we’re a public company. We’re better at forecasting. We’re better at planning. We’re better at executing.”The trap, he says, is when you let financial discipline replace strategy instead of complementing it. You still have to invest aggressively in AI and new products. You still have to compete. You still have to go into new areas — Atlassian just shipped a customer service product. Their service collection business could go public by itself.The private companies have the luxury of not marginal-costing every decision. Their CMOs don’t have “pesky public market investors” asking about profitability. That creates a real asymmetry. But Mike argues the discipline of public markets is actually a long-term advantage — you just can’t let it make you timid.7. If You Don’t Enjoy It Anymore, That’s Okay — But Be Honest About ItMike shared something that really resonated. His co-founder Scott Farquhar retired a year and a half ago after 23 years. Mike starts work at 5 AM every day now. He’s working harder than ever.And his advice to the SaaS founders he’s been counseling (he called it “SaaS therapy”) was refreshingly direct:First, accept reality. Stop pontificating about whether AI is going to change things. It already has. Go build something.Second, ask yourself honestly: do you still enjoy this? Not “do I enjoy every minute of every day” — that’s not realistic. But over a 90-day average, over a year, would you choose this job again today?If the answer is no, that’s fine. You built something amazing. Hand it off thoughtfully and go do something you enjoy. That’s not weakness — it takes more courage to step away than to hang on collecting a paycheck until the board forces you out.If the answer is yes, then get to work. We’re going to create our way out of this, not hide in a hole and wait for it to pass.CEO for 24 Years in the Age of AIMike Cannon-Brookes has been building Atlassian for 24 years. He’s seen every “this time it’s different” moment the tech industry has produced. And his message is clear:B2B software isn’t dead. It’s changing. The companies that will win are the ones that accept reality, invest aggressively in AI, maintain execution discipline, and — critically — are actually good at what they do.If you’re a founder reading this and feeling anxious about the AI transition, here’s your to-do list:* Figure out if you’re selling to input-constrained or output-constrained functions* Stop reading “SaaS is dead” takes and go ship AI features your customers actually want* Make sure your inference costs are going down, not up* Ask yourself honestly if you still want to do this* If yes, start at 5 AM tomorrowThe era of autopilot SaaS growth is over. But if you have 350,000 customers, great distribution, engineers who can build, and a product people love? That’s a pretty good starting point to build a company.You just have to be good.Want to learn how to WIN in the AI Era in B2B? Join 10,000 of us at SaaStr AI 2026 May 12-14. We'll give you the playbooks to win in AI + B2B in 2026. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

High inference costs are OK—if they make your product so viral and so competitive it almost sells itselfHere’s the counterintuitive insight that’s reshaping how the smartest AI founders think about unit economics:Your inference costs aren’t your gross margin problem. They’re your CAC replacement.The companies growing fastest right now—Cursor crossing $1B ARR with ~300 employees and no traditional marketing, Lovable hitting $300M ARR with zero paid acquisition—aren’t sweating inference costs. They’re leaning into them. They’re treating compute as their primary growth investment, not their primary margin drag.This is a fundamental reframe. And if you’re still optimizing for gross margin while your AI-native competitors are optimizing for virality, you’re playing the wrong game.The Math That Traditional B2B and SaaS Gets WrongOn a recent 20VC x SaaStr episode, we discussed Anthropic’s inference costs coming in 23% higher than expected. My immediate reaction was pessimistic for mid-market B2B SaaS:“I worry this is the final nail in the coffin. You did everything right—got profitable, built an agent—and now you just can’t afford the inference to compete.”Here’s the scenario: You’re a $50M ARR B2B company. You built the agent your board demanded. Your agent costs $2.50 per interaction. You need 50 million interactions to stay competitive. That’s $125 million in inference costs on $50M in revenue.Game over, right?Not necessarily. The question isn’t whether you can afford the inference. It’s whether the inference makes your product so good that sales and marketing become irrelevant.The Cursor Playbook: Inference as DistributionCursor crossed $1B ARR by late 2025—roughly 24 months from launch—with about 300 employees and minimal traditional marketing. They went from $100M ARR in January 2025 to $500M by June to $1B+ by November. The fastest SaaS growth curve ever recorded.How? They spent aggressively on inference to create what Andrej Karpathy called the “vibe coding” experience—the moment when developers forget they’re writing code and just describe what they want. That experience is computationally expensive. It requires reasoning tokens, multiple model calls, context management across entire codebases.Traditional SaaS math would call this margin suicide. But here’s what actually happened:* The “wow moment” converted instantly. Developers tried Cursor, experienced something magical, and became evangelists within hours.* User-generated content became their entire marketing funnel. Every tweet about “I built an app in a day with Cursor” was free distribution worth thousands in CAC.* The viral loop compounded. Engineers at OpenAI, Midjourney, Shopify, and Instacart started spreading it organically. No sales team required.* Conversion was frictionless. $20/month is an impulse buy when the product makes you demonstrably faster.The inference spend wasn’t a cost center. It was the marketing budget. It just showed up on a different line item.Lovable’s Rocket Ship to $300MLovable hit $300M ARR in January 2026—roughly 14 months after launch—with fewer than 200 employees and zero paid acquisition. That’s still $1.5M+ revenue per employee, nearly 8x the industry benchmark.Their secret? They engineered virality into the product itself. When users build apps with Lovable, the outputs are shareable. The AI-generated code is good enough that users want to show it off. Every app becomes a piece of marketing collateral.The underlying inference cost to generate these apps is significant. But look at what they avoided:* No enterprise sales team (zero)* No paid acquisition (zero)* No SDRs cold-calling (zero)* No expensive conference sponsorships (zero)The inference is the go-to-market motion. The product is the marketing.You Can’t Have It Both WaysHere’s the brutal math that too many founders are ignoring:You can’t have high inference costs AND high sales & marketing costs. At least not for long. It has to come from somewhere.Traditional SaaS could absorb 40-50% S&M spend because gross margins were 80%+. There was room. The unit economics worked.But when your gross margin drops to 50-60% because of inference costs, that room disappears. You’re now choosing between two paths:Path A: Inference-First (Cursor, Lovable)* Gross margin: 50-60%* S&M: * Growth driver: Product virality* Requires: Magical product that sells itselfPath B: Sales-First (Traditional Enterprise SaaS)* Gross margin: 75-80%* S&M: 40-50%* Growth driver: Sales efficiency* Requires: Lower inference costs, less AI magicWhat you cannot do is run 50% gross margins AND 40% S&M. That’s negative operating margin before you pay a single engineer. That’s burning cash with no path to profitability. That’s a company that dies.The trap I see founders falling into: they build an AI product with significant inference costs, then layer a traditional enterprise sales motion on top of it because “that’s how you sell to enterprises.” Now they’re paying for compute AND paying for a sales team AND paying for marketing—and wondering why their burn rate is out of control.Pick a lane. Either your inference spend IS your sales & marketing (because the product is so good it creates its own distribution), or you optimize inference costs aggressively and invest in traditional go-to-market.The companies trying to do both are the ones that will run out of runway.The Third Path: Build Something So Valuable They’ll Pay $50-100K+ (More)There’s another way to solve the inference cost problem that doesn’t require virality or zero sales spend.Build an AI agent so valuable that businesses will gladly pay $50,000 to $100,000+ for it. Even small businesses.This isn’t fantasy. I talked to a sales leader at an agentic AI company recently. I asked him how many figures are in his mid-market deals. I was thinking $20K to $50K, maybe $75K on the high end.He said they’re all seven figures.Seven figures. For mid-market.When you price based on labor replacement instead of software seats, the math completely changes. You’re not competing with Salesforce’s $150/seat. You’re competing with a $120K SDR salary, a $180K AE salary, a $95K customer support rep.The current market rate for agentic GTM tools: $100K+ per agent as a starting point. Some companies are paying $50-70K just to get started, plus another $25K for a forward-deployed engineer to set things up.And here’s the thing: smaller businesses will pay this too if the ROI is obvious enough.Think about it from the buyer’s perspective:* Human SDR loaded cost: $100K-$130K annually* AI SDR: $10K-$15K for basic, or $50K-$100K for enterprise-grade with support* Even at $100K, that’s still cheaper than a human—and the AI never has a bad day, never forgets the product, never needs ramp timeAt $50K-$100K per agent, your inference costs become a rounding error. You’re not optimizing tokens. You’re delivering ROI measured in weeks, not quarters.This is the premium escape hatch from the inference cost trap: Don’t try to win on efficiency. Win on value so overwhelming that price becomes irrelevant.The Inference Squeeze on Mid-Market B2BHere’s where it gets brutal for traditional SaaS.If you’re a $50-100M ARR company that spent years building a sales org, optimizing your marketing funnel, and perfecting your enterprise motion—you now face a new type of competitor. One that has no sales team to pay, treats inference as their primary distribution channel, and can invest every marginal dollar into making the product more magical.You can’t out-spend them on ads because they don’t run ads. You can’t out-execute them on sales because they don’t have salespeople to compete against. Your moat was go-to-market efficiency. Their moat is product virality powered by unlimited compute.The existential question becomes: Can you afford NOT to spend on inference?The Framework: Inference as CACHere’s how to think about this:Traditional SaaS Model:* Gross margin: 75-80%* S&M spend: 30-50% of revenue* CAC payback: 12-24 months* Growth driver: Sales & marketing efficiencyAI-Native Model:* Gross margin: 40-60% (temporarily)* S&M spend: * CAC payback: Near-instant (product is the acquisition)* Growth driver: Product virality via inference qualityThe question isn’t “What’s your gross margin?” The question is “What’s your blended customer acquisition cost when you include inference spend that drives organic growth?”If spending an extra $1M on inference generates 10x that in organic revenue—because your product creates its own distribution—you should spend that million dollars every time. Even if it temporarily crushes your gross margin.The New Competitive MoatThe winners in the AI era won’t be the companies with the best gross margins. They’ll be the companies that figured out how to turn inference into distribution.This requires:* A product with a shareable “wow moment.” Not just good—magical. The kind of thing users can’t help but tweet about.* Viral mechanics built into the core experience. Every output should be a potential piece of marketing content.* Pricing that makes conversion frictionless. $20-50/month that’s an immediate no-brainer when the value is obvious.* The financial runway to absorb inference costs while the viral loop compounds. This is where well-capitalized AI startups have an edge—Open Evidence, Harvey, Lovable all raised massive rounds specifically to fund this strategy.The Hard Truth for IncumbentsHere’s what I worry about for traditional B2B:You did everything your board asked. You got profitable. You built an agent. It works. Your customers like it.But your agent costs $2.50 per interaction. You need 50 million interactions this year. That’s $125M in inference costs.Meanwhile, your AI-native competitor raised $300M specifically to fund inference spend. They have no sales team. Their product is so good it creates its own distribution. They’re growing 300% year-over-year while your growth is decelerating.If you walked into your board meeting and said “Good news guys, inference costs are going down 30% this year because our IT team’s really good at managing costs,” I would throw my mouse at the monitor.The only way out is to build an agent so good you can charge $10-20K per month because it replaces 20 people with ROI measured in weeks—not a sales pitch claiming it’s that good, but literally that good.Or: accept that inference isn’t your enemy. It’s the new battlefield. And the companies willing to spend aggressively on compute to create magical, viral products will win—even if their gross margins look terrible by traditional SaaS standards.Inference Is The New Sales & Marketing. At Least, a Lot of It.The AI companies winning right now have flipped the script entirely:* Cursor: $1B+ ARR, minimal marketing, inference is the product that sells itself* Lovable: $300M ARR in ~14 months, zero paid acquisition, virality is the distributionThey’re not worried about inference costs crushing their margins. They’re worried about inference costs being too low to create competitive differentiation.That’s the paradigm shift. Inference isn’t the new COGS. Inference is the new S&M.Spend accordingly.Three Takeaways for Founders:* Reframe inference as acquisition cost, not gross margin drag. If spending $1 on inference generates $10 in organic revenue through virality, that’s a 10x return on your “marketing” spend.* Build shareable “wow moments” first, optimize margins later. The companies growing fastest right now prioritized magical user experiences over unit economics—knowing the economics would follow.* If you can’t afford to compete on inference, find a different wedge. Either charge enough that inference costs don’t matter ($10K+/month with clear ROI), or accept that you’re competing against companies with infinite compute budgets and no sales teams to support.Thanks for reading SaaStr AI: How To Sell, Scale, and Win! Subscribe for free to receive new posts and support my work. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

There’s a growing wave of AI agent skepticism on LinkedIn right now. And some of it is earned. A lot of founders bought an AI SDR, didn’t train it, and got garbage results. Then they posted about how “AI agents don’t work.”But here’s what we know after 8 months of running 20+ agents across our entire go-to-market at SaaStr — with just 3 humans and a dog: $4.8 million in additional pipeline sourced by agents. $2.4 million in closed-won revenue. Deal volume more than doubled. Win rates nearly doubled. And none of it cannibalized our existing inbound.It works. But not the way most people think it does.Let me break down what we’ve actually learned — the real stuff you won’t see in the LinkedIn posts.The Results Are Real, But So Is the WorkLet me give you the honest numbers first.Eight months in, our AI agents have generated $4.8M in additional pipeline and $2.4M in closed-won revenue that was first-touch sourced from an agent. Our deal volume has more than doubled. Our win rates have nearly doubled. And we’ve sent over 60,000 high-quality AI-generated emails just on the sales side — not even counting the nearly 1 million interactions through our vibe-coded apps.Here’s what matters most about those numbers: this was all additive. It did not cannibalize our other inbound revenue sources. We didn’t drop anything when we deployed these agents. We still send marketing emails. We still do outbound ourselves. We still send gifts. We still invite people to the SaaStr house. All the things we used to do before — we still do them. The agents augmented everything.But here’s the honest truth you won’t see on LinkedIn or X: we maintain these agents every single day. Literally every morning before anything else, we’re checking our agents. Amelia and I each spend 15-20 hours per week — that’s each, not combined — actively managing, iterating, checking responses, making sure nothing hallucinates, making sure the agents are talking to people the way we want them to.The time we used to spend managing humans on our team? We now spend that same amount of time — if not more — managing agents. The difference is there’s no people drama, and the agents work at a much higher capacity and scale than a human ever could.At some point, you realize you simply cannot keep up with your agents. They’re faster than you. They work 24/7/365. They can always answer a question, always book a meeting, always reach back out. The humans become the bottleneck.The Secret Nobody Tells You: Agents That Require Deep Training Cannot Be Self-TrainedI was meeting recently with the CEO of a next-generation AI go-to-market company — they already have millions in revenue and are publicly launching soon. I asked what their secret sauce was.The answer: they do everything. The onboarding, the tagging, the first campaigns — all of it. They do it almost to a fault. Some customers think it’s too easy and don’t even realize how much human energy is going into deployment behind the scenes.That’s the learning. If you haven’t deployed many agents — or any for real — you need to have an honest conversation. Not with someone in sales who doesn’t know how the product works. Talk to a forward-deployed engineer. Talk to a leader. Find out what it’s actually going to take in the first 14 days, the first 30 days, and every single day after that.Then you have to actually do it. Otherwise, it’s like going to the doctor, getting a prescription, and never taking the medicine. It literally will not work.A lot of the agents we use are pushing downmarket to be more self-service. So far, that doesn’t work. Agents that require deep training cannot be self-trained yet. It will come — agents are getting dramatically better every quarter. But for now, be skeptical. If you buy a cheap tool that claims it’s self-trained, make sure it actually works. If you buy a more complex tool, talk to someone senior enough on deployment who actually knows the product.The 90/10 Rule: Buy 90%, Build Only 10%Here’s our rule of thumb: buy 90% of your AI stack. Only build the 10% where no vendor can do it well and it’s a P1 priority.We’ve followed this ourselves. The vast majority of our agents are third-party tools that we’ve trained and customized heavily. We only built custom agents where we had a very specific use case that no vendor could handle — like our AI VP of Marketing (more on that below).Kyle, the CRO at Owner (one of our portfolio companies), has followed roughly the same approach. He bought a bunch of third-party agents, made them work, and then hired a former founder/engineer — someone who was literally a CEO of an LLM company — to build a proprietary in-house tool for the 10% that needed to be custom.That’s an extreme case. For most of you, building custom agents probably won’t make sense yet. Focus on making the bought tools work first.How to Evaluate AI Agent Vendors (Don’t Skip the Basics)I don’t know why people throw away basic evaluation practices just because something has “AI” in the name. Here’s what we do — and what you should do:Ask for a dedicated resource. I asked every single AI tool we now deploy for help. I told them: one, I’m going to need an FDE (forward-deployed engineer), and two, let me talk to people who have actually used this.Ask for customer references. I see too many people skip this because “it’s an AI tool.” Ask for a reference. Ask for one in your vertical if possible. If they push back, maybe don’t use that vendor. Most of these companies have at least one customer that’s somewhat like you.Demand FDE time at deployment. Salesforce put an FDE on our success team. Qualified has an FDE on our success team. Replit has an FDE on our success team. With Artisan, anytime I have an issue or idea, I go straight to the CEO or head of product. You should ask for — and expect — some level of dedicated support to get started. You may not need it weekly later, but you absolutely need it upfront.Trust your gut. If it doesn’t feel right, don’t buy it. If your Spidey sense says this agent isn’t going to work, it won’t work. Buy another one. Even if the brand is less well-known, even if it’s scrappier — if it feels right and the team is proud to show you their product, go with that one.One more thing on evaluation: the best agents should take you as far down the journey as possible before you have to pay. A lot of AI go-to-market tools can’t practically offer free deployments due to the human onboarding costs. But the best ones get you as close to production as they can. Marc Benioff said on 20VC that he wished Salesforce had enough FDEs to get every customer into production on Agentforce before they had to pay. It’s not practical, but that’s the right instinct.Multi-Agent Management Is Messy (And That’s Okay for Now)Here’s what people always ask me: “You have 20 agents — what are you using as your MCP?”The honest answer: we don’t have one. Not a real one, anyway.What we have is what I’d call “MCP light” — a combination of Zapier webhooks, Salesforce as our system of record, and a lot of manual context-sharing between agents. We have so many webhooks firing into our Zapier account I can’t even count them. All of our third-party tools push data back to Salesforce, either through native integrations, APIs, or Zaps.Sometimes I just copy-paste context from one agent and put it into another. It’s not elegant. It’s not clean. But it works.This is the reality of multi-agent management in early 2026. It’s a lot of hodgepodging things together. But I think this is a “right now” problem, not a “forever” problem. By the second half of 2026, I believe native integrations will solve most of this.My advice: Pick one source of truth for your data (Salesforce, HubSpot, whatever) and push everything back to it. Get used to your agents talking to each other — it happens, and it’s fine. Get used to talking to your agents yourself. And for now, get comfortable with some manual context-sharing between systems.If you use specialized tools like we do (versus an all-in-one agent builder), you’ll deal with more of this messiness. We prefer specialized tools because the output quality is better. But you might reasonably trade some quality for quality-of-life by using an all-in-one platform. Both approaches are valid.A Practical Go-To-Market Flow You Can CopyHere’s an actual Zapier flow we run that you could adapt:* Catch a webhook from a website form submission* Push to Google Sheets (yes, I keep backups of everything in Sheets — don’t judge)* Create/update a contact in Salesforce* Add the contact to a Salesforce campaign (which can trigger Agentforce if you want)* Find account-level records in Salesforce to see what this company has done with you before* Enrich with Clay — pull LinkedIn activity, summarize context* Send a Slack notification with all the context: contact info, account history, Clay enrichment* Optionally generate a Gamma presentation or landing page customized for this prospect* Draft a Gmail or push to your AI SDR platformThat’s nine steps, touching six or seven different tools. It’s not simple. But it gives you a fully enriched, context-rich go-to-market motion that would be impossible to do manually at scale.The AI SDR Playbook: Hyper-Segmentation Is EverythingIf you’re rolling out your first AI SDR, here’s the single most important thing I’ve learned: hyper-segment everything.I see people running one campaign for 10,000 leads. That’s insane. I max each campaign at 100-500 contacts. Every campaign, every sub-agent gets highly customized training for the exact segment it’s targeting.Don’t segment the old-school way — by geography, title, or role. None of those exist in my outbound funnel. Instead, segment by context and intent:* Website visitors (deanonymized)* Inbound leads who filled out a form* Abandoned trials or carts* Event leads from webinars or conferences* Former customers who changed jobs* Current customers for expansion* Recent marketing leads from gated content* Leads you never followed up with (we famously gave these to Agentforce)* Lapsed contacts you forgot to talk to* Low-scoring leads that still show intent but nobody wants to call back* Alumni from previous events* Warm referrals and community membersI haven’t exhausted this list after 8 months. Start here, not with cold outbound. Your AI agent has zero context for why it should reach out to a random cold lead. But it has rich context for every segment above.The more context you give your agent, the better the output. This is no different than how you use Claude or ChatGPT every day — you talk to it, give it context, tell it about your business. Same rules apply with AI SDR agents.And tell your agent what you can’t do. This is a nuance I only learned after months of deployment. AI agents are self-gratifying — they try to beat their own metrics. Over time, they start making promises you can’t keep. Ours started offering people speaking slots at SaaStr Annual, which isn’t how our speaker selection works at all. Once I explicitly told the agent what we don’t do, the quality jumped significantly.Bad context equals bad emails. Period. Train on the best of everything — your best email copy, your best follow-up sequences, your best case studies — and then also define the boundaries clearly.Why We Built Our Own AI VP of MarketingWe looked at every third-party AI marketing tool on the market. None of them could do more than content. The real problem we needed solved was orchestration — coordinating data across agents, prioritizing initiatives, and creating executable daily plans grounded in actual performance data.We also knew from experience that every time we tried to onboard a human with all our data, they got overwhelmed. So we built our own AI VP of Marketing. We nicknamed it “10K” — because the goals were 10,000 attendees at SaaStr Annual and $10M in revenue for the year.Here’s how we built it:* Collected data from our agents, third-party tools, Salesforce, Zapier workflows, and internal historical data* Fed it into Claude Opus with very clear goals and context (this required upgrading to Max — Pro ran out of memory over a weekend of heavy analysis)* Pushed the output to Replit so the whole team could access it as a web app* Told it to generate both high-level strategy and granular daily executable tasksThe key instruction: “Don’t give me generic strategy ideas. Give me executable tasks grounded in our data that three humans and a dog can actually do.”What came back was remarkable. A week-by-week game plan with specific campaigns, specific timing, specific channel allocation. It told us how much to spend on LinkedIn ads and what the ads should say. It told us which campaigns to run through Artisan vs. Qualified vs. Agentforce. It literally told me what to post on social media.Some stuff it said to blow up or abandon. Some stuff it said to bring back. And a bunch was net-new. Because it was all rooted in data, I trust it.I talk to 10K every day. “Where are we? What should we be doing today? Where are we falling behind?” It keeps me honest and focused. Sometimes I push back — it once suggested a campaign I didn’t think was compelling enough. We debated it, it looked at the data, and agreed to change course.It’s not always right. But it’s not always wrong either. And it keeps me more organized in this one vector than I’ve ever been.The Maturity Curve Is RealHere’s the honest assessment of where AI tools stand today:Coding and support tools → Most mature. These work well today.Sales tools (AI SDRs, conversational agents) → Getting there. Work well with significant training and maintenance.Marketing tools → Not nearly as mature as vendors claim. That’s why we had to build our own AI VP of Marketing. Most marketing AI tools only handle content. Orchestration, campaign planning, cross-channel coordination — none of that exists in a turnkey product yet.By the second half of 2026, I believe all of this will connect natively. Your AI VP of Marketing will integrate directly with your AI SDRs, your Agentforce, your Clay tables, your LinkedIn ads — all of it. There won’t be any excuses for shooting from the hip in B2B marketing anymore.But we’re not there yet. For now, it’s messy, it’s manual in places, and it takes real human effort every single day.The Bottom LineIf I had to distill everything we’ve learned into one piece of advice: find something in your go-to-market motion that just isn’t getting done, or is getting done at a mediocre level, and put an agent on it.Don’t try to replace what’s already working well. Do that as your 10th or 20th agent. Start with the low-hanging fruit — the customers that are too small for your team to call back, the leads that take too long to respond, the contacts with lower scores that still have intent but nobody wants to prioritize.Because even modest yield from those segments is magical. It’s entirely additive revenue that you never would have captured otherwise.The emails our agents send aren’t the best ever written. I’d say they’re “pretty good.” But we’re getting scale — 60,000 high-quality, personalized emails that we could never have done manually. More high-quality, pretty-good interactions with more people, more often. That’s the formula.And that’s worth $4.8 million in pipeline and counting. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

If Growth Isn’t Accelerating, You’re Not an AI Company. And 9 Other Hard Truths for SaaS in 2026.I had a great conversation with the TBPN crew the other day, and we covered a lot of ground — from the state of the SaaS market to PE exits to vibe coding to how agents are already reshaping how software gets bought and sold. I wanted to pull together the key themes here, because I think founders need to hear some of this, even if it’s uncomfortable.Let me walk through the big takeaways.1. If Growth Isn’t Re-Accelerating, You’re Not Really an AI CompanyThis is my simple rule, and I think it cuts through all the noise.Every public company, every startup, everyone is talking about their “AI strategy.” They’ve built an agent. They’ve shipped a copilot. They’ve got an AI tab on their website. Great. But has growth actually re-accelerated?That’s the bull case for Meta. They genuinely accelerated growth. The AI they baked into their ad-matching platform is working. Retail advertisers generating ads with AI tools — it’s all compounding. It’s real.Now look at the flip side. Microsoft’s AI business is still blowing up, but they missed on the software side. The Trade Desk has been destroyed. Figma is trading below 10x revenue despite essentially creating and owning a category.The point is: AI talk is cheap. Revenue acceleration is the only metric that matters now. And I’ve lost patience — with founders at $1M, at $10M, with public companies — who haven’t seen the lift. ElevenLabs just crossed $350M. MongoDB dramatically re-accelerated. Show me the money.2. The Transition from “Deeply Tough Love” to Just “Tough”Last year, my advice to founders was deeply tough love. The market was brutal, and most people hadn’t adjusted.Now it’s just tough. Here’s why: you’ve had time. Claude got really good at 3.7 — that’s why Replit and Lovable blew up. That was a year ago. Whether you’re Agentforce or one of my portfolio companies, you had a full year to re-accelerate growth. Some did. Most didn’t.And honestly? Salesforce is doing better than some startups I work with. We’re actually probably the only organization of our size using Agentforce for real, every day. It works. I can’t tell you how many startups whose “agentic product” is still basically a copilot.3. 80% of Your Team Wants to Work Like It’s 2021There’s a narrative that AI has reinvigorated SaaS founders — that people who got to growth stage a decade ago are suddenly back in the arena, fired up, tinkering with tools, pushing teams harder.It’s a great narrative. In the real world, it’s not that common.I talk to public company CEOs in B2B, my own portfolio, others — a lot. Behind the scenes, off the record. And since our agents blew up, everybody thinks we’re some kind of GTM agent gurus. So they come to us.The consistent theme: 80% of their team wants to work like it’s 2021. Everyone has to create a skunkworks team or something. And meanwhile, the AI-native companies are crushing it precisely because they don’t have to deal with 20,000 pre-AI customers who still have feature gaps, clunky software, and legacy competitors. Then you’ve got 10,000 new AI competitors who don’t carry any of that baggage.I’d love to tell you I know dozens of companies that went from 40% growth at the start of 2025 to 80% or 110%. I can think of a handful.4. The Vibe Coding Problem Is Worse Than You ThinkSix months ago, the problem was 10 companies running at every exciting category. Today it’s worse.A very successful seed investor — relatively new to the game — told me yesterday he’s giving up because everyone can vibe code something. He can’t even tell the difference between founders anymore.Now look, I’ve vibe coded 20+ apps that have been used over a million times. I’m in the top 0.1% on Replit. I know a bit about this. You’re not going to vibe code Salesforce. But you can vibe code something for Demo Day that looks really good. Stuff that 18 months ago would have made your jaw drop — “oh my god, an agent for dental follow-ups that’s fully automated!” — that’s now table stakes. I built my own version on Replit a week ago. It was already obsolete today.Here’s the real unlock though: if you want to build super-niche software for real — not for an hour, not one-shot, but actually commit to it — you can now do it without an engineer. RevenueCat, a company I was the first investor in, powers mobile subscriptions for 50% of mobile apps. Andreessen published data showing the number of new mobile apps basically quintupled at the end of last year because of vibe coding. And it’s just starting.The very bottom of the market is exploding with possibility. Salesforce isn’t going away. But that middle is going to be harder.5. Investors Want Insane Growth. And There’s No Great Answer.The challenge isn’t just the clones. It’s that investors now expect insane levels of growth. Going from $1M to $100M ARR in a year used to be almost unprecedented. Now it’s what you need to be fundable.HigsField for video is over $200M. But that’s not the one getting talked about every week — that’s Harvey, Lovable, Replit.And some great businesses will compound to solid rates but remain unfundable. They would have been unfundable two years ago too. I genuinely don’t know what the answer is for these founders. A company projecting 3x growth this year? Beautiful deck. But that growth rate is now a nightmare for fundraising because it doesn’t look best-in-class.6. PE Has Said Goodbye to B2B SaaSI know this for a thousand percent sure, and it’s slightly inconsistent with the data Carta and others put out: these pre-AI B2B companies at $50M, $100M, $200M, even $800M — no one wants to buy them.Hooray, you got profitable. That means you’re not going bankrupt. You have not solved your existential problem in AI.I’m advising a company at about $140M in revenue with decent growth. We did an M&A review recently, and I was shocked at how many companies much bigger than them are also on the block. They’ve already gone to Thoma Bravo, Vista, Insight. Those firms aren’t doing a $140M ARR cram-down combo exit unless everyone else in the PE chain already said no.Until late 2023, if you got to $20M in revenue, you were still growing, and you were efficient — someone was going to buy you. The question was whether it was 5x, 6x, or 10x. That playbook is dead for PE and it’s dead for exits.I do think the Thoma Bravos of the world will figure out how to accelerate AI in their portfolios. If you can own the agents on your platform, that’s how you re-accelerate. But if you don’t, the agents will take away all the value. And I’ve lost patience with founders in la-la land about this.7. How We Went from 8 Salespeople to 1 Human + AgentsWe’re running four different sales agents right now: Agentforce, Artisan (hot YC company), Qualified (Salesforce just bought them for almost a billion), and Clay (just raised at $5B valuation). All of them, for different use cases. It’s work. But it works.The specific use case where Agentforce shines for us: reactivations. Folks who used to be sponsors or attend our events that a human was simply too lazy or too busy to follow up with. We scored them, sent them out with the agent. 70% open rate.We went from eight people on our sales team to one human plus AIs. The agents do the work humans gave up on — looking up who replaced the person at Ramp or Brex, following up without shame, closing deals on Saturday night.An agent closed a $100K deal on a Saturday night. How many humans want to do that? They’re streaming Netflix. The agent doesn’t quit, doesn’t take vacations, doesn’t need onboarding, and doesn’t demand a raise six months in.That’s why last June, Amelia and I just said: we’re going all-in on agents. We’re going to push every agent to its limit. Now three people are doing the work of 15, and we replaced two agencies with apps we built ourselves.8. GEO Is Real, But Deeper Than People ThinkPeople are underestimating how LLMs are reshaping software discovery. “GEO” — generative engine optimization — is nice as a concept, but I think it undersells what’s actually happening.I did this experiment yesterday: went into Replit and asked, “What’s the best CRM for me to use?” It said HubSpot. That’s where the money is.We’re not going to Google anymore for this stuff. We’re asking the agent. Or even more interesting — we’re not even asking. You say “I need a website” and the agent picks a database for you. You don’t even know what it picked.I built a digital JSON tool that’s been used 175,000 times where people just ask “what should I use?” Why would you go to Google and pay? Why sit through endless webinars with a sales rep who doesn’t know the product?Look at Resend vs. SendGrid. I couldn’t get SendGrid to work in Replit — the founders are long gone, they throttled the free tier into uselessness. The agent said “use Resend.” I’ve never gone back. WorkOS was around for years trying to do OAuth, going through layoffs, going nowhere. Then every agent started recommending it and the company blew up.This is still nerdy-early, but some version of this is how everyone will pick software. The companies that win agent-driven discovery win the next decade.9. The Only Niche AI Investments Worth Making: Show Me the 10x PricingWill a lot of niche AI-powered software blossom? Absolutely. Dairy farms, hydroponic operations, spa management — all great. But to make money investing in these, the question is: can the agent deliver so much ROI that you can charge 4x, 5x, 10x more than the previous category leader?If it’s the same unit economics as deals I looked at five years ago, I’m out. Even getting to $100M is tough.But look at the AI SDR companies — Artisan, Qualified, Clay. They’re charging $100K+ to start. The last generation (SalesLoft, $2.3B exit to Vista in December 2021) struggled to get customers to pay $100K. Now it’s table stakes.Mangomint, a company I invested in, does software for spas and salons — a crappy, hyper-competitive category with 10 solid incumbents. After they hit $30M, they blew up with agentic and automation features. Not because it made the product marginally better, but because it made the product dramatically more valuable. An $8K/year product that can now credibly charge $80K because it genuinely eliminated ten back-office positions? That’s the math I want to see.10. Don’t Quit Your JobOne last thing. I was at an event this week with a lot of C-level B2B executives. I’ve gone every year for a while.This year, they finally admitted it: they can’t find jobs. People with 2021-2024 tech tool skill sets — nobody needs them anymore. It’s finally sinking in.So whatever you do, don’t quit. If you like it at your company — stay. The wealth creation happening at the top (20,000+ NVIDIA employees have made $20M+ in the Bay Area alone) is about to explode further. The best VCs and best engineers will make money like we’ve never seen before.But the middle is going to have a very rough ride.The IPO market is open-ish but discriminatory. EquipmentShare crushed it. Wealthfront bombed. We’re entering an era of wealth creation from IPOs unlike anything we’ve seen — trillion-dollar exits are becoming conversational, not aspirational.Just make sure you’re on the right side of it. Build something real, show the revenue lift, own your agents, and don’t wait for PE to save you. They’re not coming. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

A SaaStr AI deep dive with Sam Gutmann, CEO of Own, on building a billion-dollar backup company by saying “no” to almost everything. He joined Harpinder Singh (Partner, Innovation Endeavors) to share the whole story — and his top mistakes.And come hear 200+ stories like this at SaaStr AI Annual May 12-14 in SF Bay!!Top 5 Takeaways1. The CEO who dismissed “Backup for Salesforce” as “the dumbest idea I’ve ever heard” went on to build the category leader.Sam literally stopped a board meeting to call out how stupid he thought the idea was. Six years later, he was running the company that would define the space. Markets evolve. Your priors can be wrong. The best founders update their views when the data changes.2. Don’t expand until you cross $100M ARR if your core market is still only single-digit penetrated.Own had backup for ServiceNow, Microsoft, and other platforms ready to go for years. They said no. They killed products that weren’t generating revenue. The result? 100%+ annual growth rates by staying focused on Salesforce until they had the resources to truly do multi-platform right.3. The CEO ran the financial model himself until $200M ARR — and that’s why they hit their numbers.When their outsourced CFO offered to run FP&A, Sam said “absolutely not.” Every investment tied back to a cell in his Excel model. The outsourced finance firm told him: “You’re the only founder where our FP&A team isn’t doing this for you. You’re also the only company actually making their numbers.”4. “Ideas are worthless. It’s all about execution.”Salesforce came out with a competing product. It didn’t work. They killed it. They tried again. It still didn’t work. Then they acquired Own. When you have 1,000 people waking up every day focused on being the best backup product in the ecosystem, the platform vendor with 150 other products to sell can’t match your focus.5. The hardest leadership decision — replacing a founder or key leader who got you here — always takes too long.At a CEO roundtable, every leader agreed: firing a founder or key leader is gut-wrenching. Then they asked who would have made that call six months earlier. Every hand went up. It’s always the right decision. It always takes too long.The Origin Story: A Vacation That Changed EverythingThe story starts in the most unlikely way possible.Sam Gutmann was on vacation in Israel in 2014. He had zero network there. But he remembered that a former colleague who’d worked at the venture fund that invested in his first company had quit his job, traveled the world, and landed in Israel.“Let’s catch up over a beer,” Sam said. “By the way, I’m at a venture fund now. Know any startups I can meet?”His friend Ori said: “Yeah, I’m actually job hunting. I’m meeting with these two guys who started a backup company. Want to tag along to my unofficial job interview?”Sam had the tour van driver pull over in a city he’d never heard of called Herzliya. They sat down at a coffee shop with Ariel (the founding CTO) and two friends who’d started a part-time project called OwnBackup.Halfway through the meeting, they turned to Ori and said, “Please stop selling yourself. We’re not hiring a sales guy.”Then they turned to Sam: “We actually are hiring a CEO. Are you interested?”That one-hour coffee ruined the rest of his vacation. But ten years later, they sold to Salesforce.The Irony: Sam Thought This Was a Terrible IdeaHere’s the thing that makes this story remarkable.In 2008, Sam was in a board meeting for his first company — an online backup service where software ran on servers and sent encrypted data to the cloud. They were brainstorming growth ideas.The chairman walked up to the whiteboard and wrote in red marker: “Backup for Salesforce.”Sam stopped the meeting.“That’s the dumbest idea I’ve ever heard.”Six years later, he was CEO of the company that would define that exact category.What changed his mind? Simple math and pattern recognition:* Salesforce had 250,000 customers. Every single one should have a data protection strategy.* The average enterprise uses 300+ B2B applications. The same vulnerability exists whether your data is on a laptop, an AWS instance, or in Salesforce.* Every B2B provider uses the same language: “shared responsibility model.” They’re responsible for the platform. You’re responsible for your own data.Sam likens it to an apartment building: the landlord handles the infrastructure, windows, pipes, and elevators. But you’re responsible for everything inside your unit.The Path to Product-Market FitThe founders of Own had an unusual origin story themselves. They ran a traditional disaster recovery lab — the kind where you bring in your water-damaged phone or server with a failed RAID array, and they recover your data.Around 2010, customers started coming in saying: “You’ve recovered data from my devices before, but now I’ve lost something in the cloud. Facebook shut my account down. I permanently deleted something in Gmail. Can you help?”They couldn’t. They didn’t have access to Facebook or Google servers.After enough of those conversations, they started the first cloud-to-cloud backup in 2010. For two and a half years, it was a part-time project.Then, about a month before Sam arrived, they won a large enterprise customer in the U.S.That changed everything. Now it was: let’s spin this out of the services company, focus full-time, and find a management team that’s done this before.When Sam joined, they were close to product-market fit. He spent the rest of 2014 pressure-testing it. He installed Own on his family business’s Salesforce instance. He installed competitors. He deleted data. He tested whether they could actually recover it.His framework for evaluating the opportunity:* Is the market huge? Yes — 250,000+ Salesforce customers, plus every other SaaS platform.* Is it a real problem? Yes — his friend Ori had personally lost data in Salesforce while consulting.* Does the solution work? Yes — they proved it through testing.* Can we build a company here? Yes — not just a product, but a real business.The Power of Focus: How Own Got to $100M Before ExpandingWhen Sam joined, Own had products for multiple platforms: Facebook, Google, Twitter, LinkedIn, Gmail, and Salesforce. They were also close to releasing backup for ServiceNow.His first decision as CEO: kill everything except the product generating money.“It wasn’t until we crossed $100 million ARR that we decided we had enough resources to actually expand.” This wasn’t easy. People constantly came to him with ideas for new ecosystems. The answer was always no.“One of the hardest things to do is say no. People go, ‘I have an idea. We want to do this ecosystem.’ No, let’s keep focusing on what we do best because it’s working.”His logic was simple: when you’re growing 100% a year and the market is only single-digit penetrated, why dilute yourself? Why dilute the team? Why spend money in places where you don’t know the ROI when you have a proven market with massive headroom?The metric he used to determine when to expand: $100M ARR.At that point, the flywheel was running. He had freedom to think longer-term. He could start planning what the business should look like two to three years out.Building the Salesforce RelationshipIf you’re building in and around an ecosystem, someone’s full-time job should be managing that relationship.Own hired someone early whose entire job was building the alliance between Own and Salesforce. Not the CEO having an occasional meeting and dropping the ball on follow-up. A dedicated person.“If you’re building in and around an ecosystem, it’s really important that it’s someone’s full-time job. You really need to invest in these ecosystems.”Was he worried about dancing with the gorilla?Every board meeting, every investor meeting, the question came up: what if Salesforce competes with you?Sam’s answer: “It was not one of the top 10 things I worried about every day when I woke up.”His reasoning:* Salesforce AEs have 150+ products to sell. They were never going to focus on backup as well as Own could.* Salesforce did release a competing product. It didn’t work. They took it off the market.* They tried again. It still didn’t work.* Eventually, they decided to acquire the best technology.“Ideas are worthless. It’s all about execution. We got up every day and a thousand people were trying to execute. How do we build the greatest backup product in the ecosystem?”The Metrics-Obsessed Approach That Made Own Hit Its NumbersSam ran the financial model himself from day one.“I’ll never forget this. My first trip to Israel when I officially joined the company, I had terrible jet lag. I took a monitor back to my hotel room and I was doing this gigantic Excel model — down to how much we’re going to spend on coffee every month.”He ran that model for years.When the outsourced CFO offered to take over FP&A and run the budget, Sam said: “Absolutely not. That’s my job. I need to understand the business. I want my leaders to understand the business.”The CFO walked away, then came back with an observation:“You’re the only founder where our FP&A team isn’t doing this for you. You’re also the only company that’s actually making their numbers.”Every investment, not just coffee, tied back to a cell in the model. It was how Sam thought about the business and how he wanted his leaders to think about it.Their board decks ran 100-200 pages by the end. The strategy was partly defensive (“If there’s enough data in there, we can just spend the entire time going through it and they don’t have time to ask questions”). But it reflected a genuine culture of measurement.After board meetings, they’d pare down the deck to fit a 45-minute town hall and share as much as possible with the entire company.“Did every SDR remember every metric? No. But that trust and transparency helps build trust with our employee base. It helps people think about the decisions they’re making better.”Culture Can’t Be Outsourced to HRSam interviewed every employee through about 400 hires. He probably should have done it through 1,000.At some point, he wasn’t interviewing for technical skills — he trusted the manager for that. He was interviewing for cultural fit.“People are working their butts off, and if they’re not enjoyable to be around and good energy, it’s really hard to scale.”His philosophy on culture:* Culture is not something HR owns. It starts at the top.* You can’t outsource it. Every decision should consider: how could this impact culture?* Every new employee should understand they’re part of the culture. It’s their job to commend colleagues who uphold values and remind people when they don’t.The reviews on Salesforce’s AppExchange tell the story. Own had five times more reviews than their closest competitor. Many weren’t just “the product works.” They said things like: “The product works, but Josh in support was awesome” or “Jane, our sales rep, wasn’t trying to sell me. I was treated as a real partner.”That’s whole product — not just technology, but everything around it.The Hardest Thing: Knowing When to Replace LeadersA sales leader who can scale from zero to $200M is the exception, not the rule.“There are great sales leaders that can go from zero to a million, or one to ten million, or ten to a hundred. That person is probably not the person that’s going to go from 100 to 200. That’s perfectly fine.”Sam’s perspective: a leader who’s great at 1-10 should do that 5-10 times in their career. They’ll be happy and make good money. If they’re stretching way beyond their comfort zone, they’re probably miserable and you’re probably not making your numbers.At a CEO roundtable, the moderator asked: “What’s the hardest challenge you have to deal with?”Almost everyone agreed: firing a founder or key leader.Then the follow-up question: “Once you’ve done it, who would have not wanted to do it six months prior?”Everyone.“It’s always the right decision. It always takes too long.”The moderator, a successful entrepreneur who’d nearly taken his company public before selling, offered perspective: “I agree. When I was running a thousand-person company, it was the hardest thing to do. But now I own a professional sports team. When I make a change to the roster, there’s millions of angry fans tweeting at me that I’m a jerk. I would die to go back to a thousand-person software company and make a change like that.”Hire People Who’ve Done It BeforeSam was always a fan of hiring people who have done it before, not just smart athletes.One of Own’s other investors put it bluntly:“Look, we gave you a lot of money to go execute as fast as you can. We didn’t give you a lot of money to go try. So why are you betting on someone that may be able to do it? Go buy the person that’s done it before two or three times.”That doesn’t work in every role. You have to consider what it would do to culture to bring someone too many steps ahead. But generally: bring someone who’s done one or two steps beyond where you are today, and you’ll get there faster.On sales leadership specifically: when you think you’ve hit product-market fit or you’re close, figuring out how to operationalize and institutionalize the sales process is critical.“It’s great to have a lone wolf sales rep that exceeds quota, but that’s not replicable.”Doing this too late means you might get a few sales early, but scaling becomes much harder.The Decision to Sell to SalesforceOwn wasn’t running an active sale process. They were talking to bankers about an IPO.But they’d known the Salesforce team for years. Salesforce Ventures had invested in their seed round alongside Innovation Endeavors and continued to support them. By the end, Own was one of Salesforce’s top five ISV partners.The conversation started as: if there’s another round, would Salesforce participate?At some point, Salesforce said: “Why don’t we just buy you?”It wasn’t 2021 anymore. Markets had changed. But Sam kept coming back to the mission: every one of those 250,000 Salesforce customers should have data protection. The fastest way to get Own’s technology into every customer’s hands was through Salesforce’s go-to-market and distribution.Was the threat of Salesforce competing a factor in the decision?“A little bit. If they wanted to make our lives miserable, they could. But this is really going to be a strategic focus for them. They’re talking about agents and AI. I don’t think they’re going to turn the organization on a dime just to compete with us.”It was one of ten factors they considered. But the primary driver was accelerating the mission while generating a strong return for early investors.Why Own Worked: Whole Product ThinkingThe product worked. That sounds obvious, but Sam remembers a competitor whose product could back up data but couldn’t restore it. (Backup without restore is just storage.)But more importantly, Own thought about whole product — not just the technology, but every touchpoint.Their reviews consistently mentioned people by name. Support was great. Sales reps weren’t pushy. Customers felt like partners, not targets.“I think it comes down to people and culture. The product works, but Josh in support was awesome. The product works, but Jane, our sales rep, wasn’t trying to sell me.”Expanding After $100M: Learning New LanguagesWhen Own finally expanded to ServiceNow and Microsoft after $100M ARR, they learned you can’t just plug-and-play from one ecosystem to another.The Salesforce ecosystem is unique. If you’ve been to Dreamforce, you know — there’s a stage where you’re learning something and stuffed animals walking around. That doesn’t exist at Microsoft or ServiceNow.The ServiceNow ecosystem speaks a different language. What customers call things is different. Own tried to apply their Salesforce playbook directly at first and had to reboot with different people who understood those ecosystems natively.“We screwed up at first when we thought we could just kind of plug and play exactly what we did in Salesforce to these other ecosystems.”The No-Substitute Lesson: Hard WorkWhen asked what to look for in a sales leader, Sam’s answer goes beyond track record and methodology.“The most important thing to look for is simply how hard do they work. There is no substitute for hard work when building a startup.”When Own was based in Israel but selling to the U.S. market, everyone at Microsoft’s startup accelerator would leave at 6-7 PM. Ghost town. Except for Ori, Own’s head of sales.“When I came to Israel, we’d work until midnight, one, two o’clock in the morning. We were the only people in the office. But we were also the only company in that cohort closing deals.”Sam’s Top 5 Mistakes (With the Benefit of Hindsight)1. Trying to plug-and-play the Salesforce playbook into new ecosystemsWhen Own finally expanded to ServiceNow and Microsoft after $100M ARR, they assumed they could apply the same go-to-market approach. Wrong. Each ecosystem speaks a different language, calls things different names, and has different buyer personas. They had to reboot with new hires who were native to those ecosystems. Should have known sooner that ecosystem expansion requires starting fresh, not copying what worked before.2. Not interviewing every employee all the way to 1,000Sam interviewed every hire through about 400 employees, focusing on cultural fit. He says he probably should have kept going to 1,000. The cultural foundation gets set early, and every hire that doesn’t fit the culture dilutes it. By the time you realize there’s cultural drift, you’ve compounded the problem across dozens of hires.3. Taking too long on difficult leadership transitionsEvery CEO at that roundtable said the same thing: when they finally made the hard call to replace a founder or key leader, they wished they’d done it six months earlier. Sam admits Own was no different. The instinct to give people more time, to hope they’ll grow into the role, to avoid the discomfort — it always costs more than it saves. The decision is always right, and it always takes too long.4. Not investing in multi-platform earlier (maybe)This one comes with a caveat: Sam isn’t sure it was a mistake. Their singular focus on Salesforce until $100M ARR drove incredible growth. But he acknowledges they might have expanded sooner given how large the other ecosystems turned out to be. The ServiceNow and Microsoft opportunities were real. The question is whether starting earlier would have accelerated or diluted them.5. Underestimating the power of in-person eventsOwn’s lead generation engine was built heavily on in-person events. Their largest investor at Insight Partners called in March 2020 with alarming news: “We’ve done an analysis of our portfolio. You’re in the top three… of companies most exposed to getting all their leads from in-person events. The world just shut down, so you’re kind of screwed.”They pivoted to online fast, but the wake-up call was harsh. A more diversified lead generation strategy would have reduced that risk. Though in fairness, the in-person approach worked extraordinarily well when events were happening.The Bottom LineOwn’s story isn’t about backup technology. It’s about focus, measurement, and execution.The market was huge. The problem was real. The solution worked. But plenty of companies have all three and fail.What made Own work:* Saying no to expansion until they’d dominated their core market* A CEO who ran the financial model himself and tied every investment to a cell in a spreadsheet* Culture that started at the top and permeated every customer touchpoint* An ecosystem partnership strategy where someone’s full-time job was building that relationship* Hard work — being the only company in the accelerator still in the office at midnight, and being the only company closing dealsWhen Salesforce finally decided they needed data protection in their platform, they had one obvious choice: the company that had spent ten years becoming the best in the world at exactly that.Sam Gutmann is the former CEO of Own (OwnBackup), which was acquired by Salesforce. He’s currently taking time off while his kids lobby for a full year sabbatical, running out of home renovation projects, and waiting for his next company to find him — ideally on vacation. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

Paul Adams is Chief Product Officer at Intercom, leading Product Management, Product Design, Data Science, and Research. He joined when Intercom was just a 14-person company after first advising the startup, and has been on the executive team ever since. Before Intercom, Paul held leadership, product, and UX roles at Facebook (Ads, Platform) and Google (Gmail, Docs, YouTube)—he was on Google’s mobile team when the iPhone launched. He’s the author of the best-selling book Grouped on social software design and co-hosts the podcast Intercom on Product with co-founder Des Traynor.When ChatGPT arrived in late 2022, Intercom was struggling—five quarters of declining revenue growth, a failed IPO attempt. The leadership team bet the entire company on AI within two weeks of ChatGPT’s release. That bet produced Fin, Intercom’s AI agent for customer service, which now resolves over 1 million customer problems per week with a 65% average resolution rate across 6,000+ customers.The Top 5 Takeaways from Intercom’s AI Transformation1. If it doesn’t feel brutal, you’re not going deep enoughPaul is blunt about this: transforming a SaaS company into a real AI company is painful. Intercom wasn’t in a great spot when ChatGPT arrived—they’d had five quarters of declining revenue growth and had abandoned an IPO process. But that pressure became an advantage.The leadership team made the call in one to two weeks. They ripped up their strategy. Ripped up their roadmap. Told the company it was happening and it wasn’t a choice.“If you’re a SaaS company who thinks you’re an AI company and you’ve not gone through brutal transformation, you’re not there yet.”The mistake most companies make? They do the easy, fun stuff—building AI features, experimenting with models, talking to customers about AI—but avoid the hard, messy decisions. Like parting ways with a third of the company because they’re not fit for the new world. Like deleting the marketing calendar and rebuilding from scratch.Paul took over two-thirds of marketing six months ago and immediately blew the entire thing up. Teams, roadmaps, calendars—gone. “The only way I knew how to build a marketing org fit for this age is to build it from scratch.”2. The only way to know if you’ve gone far enough is to go too farIntercom operates on a simple principle: the only way to find a boundary is to cross it.This shows up everywhere:Every single designer at Intercom now ships code to production. Zero did 18 months ago. The mandate was clear: this is now part of your job. If you don’t like it, find somewhere that doesn’t require it, and they’ll hire designers who love the idea.Engineering is on a path to 2x productivity—not through incremental improvements, but by declaring it non-negotiable.Paul constantly asks: “What would a brand new startup incorporated today do here?” Would they have separate product marketers and content marketers? Or is that the same job now? Would they have both product managers and product designers as distinct roles?The answer usually points to consolidation, not specialization.3. How you build software has completely changedIntercom had principles for building great SaaS products that they’d refined over years. They’d train every new designer and engineer on these principles. They were proud of them. Des had given talks about them.They had to ban all of it.The old way: Pick a job to be done → Listen to customers → Design a solution → Build and ship. Execution was certain. Technology was stable. Design was the hard part.The new way: Ask what AI makes possible → Prototype to see if you can build it reliably → Build the UX later → Ship → Learn at scale. Execution is uncertain. Design is now cheap. The hard part is the AI infrastructure—the RAG system, the custom models, the empirical evaluation.“This AI layer, our RAG system, has been 3 years in the making by a very talented team. It’s complicated. I do not understand the depths of that RAG system at all.”The visible UI is now the small part. The invisible AI infrastructure is where the real product lives. That’s a complete inversion of how SaaS products were built.4. AI products compound—every tiny improvement multipliesWhen you’re building workflows that chain multiple AI steps together, success rates multiply. If you’re at 99% accuracy on each of 10 steps, you’re at 90% overall. If you’re at 95% on each step, you’re at 60%.This is why Intercom obsesses over incremental improvements at every point in Fin’s system. They’ve run hundreds and hundreds of experiments, many of which fail. Sometimes they see an improvement in one part that degrades another.They’ve built custom models pointed at very specific customer service tasks—not general-purpose, but targeted at discrete steps in the workflow.“Each single tiny incremental improvement in each of these steps adds up to the highest performing product, adds up to something people can trust, adds up to something people will replace their humans with.”This compounds into what Des calls the “marketing overhang” problem: lots of companies can demo an AI product, but a demo isn’t a product that works at scale. Intercom has a rule that they won’t do a product launch that isn’t a real demonstration of something they know works at scale.See: Apple Intelligence, announced June 2024, still waiting on delivery in spring 2026.5. You now have two companies—and two completely different buyersIf you succeed at the transformation, you end up with a new problem: you’re running two companies.Intercom the SaaS product: Easy product domain, predictable metrics, clear differentiators, customers who know how to talk about what they need.Fin the AI product: New product domain, chaotic metrics (“Fin’s grown 300% year-over-year—is that bad?”), customers who don’t know what they need, everything changing so fast that nobody knows what the future looks like.The buyer has changed too. In the old world, Intercom sold to customer service leaders. Simple.Now the buying committee includes: the customer service leader (influential but doesn’t make the decision), a C-level executive whose job is AI transformation across the company, and an AI-fluent technical evaluator who can assess whether the product actually works.These people live in different universes. CEOs and customer service leaders don’t go to the same events, read the same things, or operate at the same zoom level. Paul does dinners with CEOs and trade shows for customer service leaders—the Venn diagram barely overlaps.Intercom’s Top 5 Mistakes You’ll Probably Make TooMistake #1: You won’t reimagine your product—you’ll just add AI to itAdding AI features to your existing product is not transformation. Fin is a completely different product from Intercom. They have almost nothing in common. You design them totally differently, think about them totally differently.“Fin usage is eating Intercom usage at times. They’re totally different products.”Mistake #2: You won’t make self-harming decisions to winYou’ll protect revenue. You won’t want to upset the board. You won’t want to upset your sales leader. You’ll avoid taking a 10% revenue hit to do the right thing for the medium or long term.But these self-harming decisions are critical. “If it doesn’t feel really painful, you’re not deep enough.”Mistake #3: You’ll dilute the vision, delay, and convince yourself you’ve done enough“Q1 is important, let’s do it in Q2.” “Let’s dial down the vision a little.” You’ll do the easy stuff and skip the hard stuff and tell yourself you’ve transformed.Big company habits and slowness will creep in. You’ll listen too much to customers who say no to AI—the ones who said “we’re going to differentiate on human service forever.” (Those customers now use Fin.)Mistake #4: You won’t refine your company to fightYou need people who are going to fight for success, fight for these outcomes. There’s tension. There’s disagreement. People fight with each other—positively fighting, not actually fighting.“If we can do it, any SaaS company can do it. There’s nothing special about us. We just decided to do it.”Mistake #5: You’ll make these mistakes and deny it to yourselfIntercom’s exec team has the most honest, soul-searching conversations—and they still do every day. Some version of an existential question that might determine whether they succeed greatly or not.“You really have to look each other in the eye, build deep relationships with your peers and colleagues.”The companies that don’t transform will slide into irrelevance slowly. New competitors will eat their lunch. Their best people will quit to work at the real AI companies. Eventually, the company will die.Three years in, AI is inevitable. Whatever you loved about B2B, SaaS and the last decade of mobile and social—it’s gone. Will you change fast enough?Thanks for reading SaaStr AI: How To Sell, Scale, and Win! Subscribe for free to receive new posts and support my work. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

Ryan Anderson, CEO of Filevine, shared their AI transformation playbook at SaaStr AI London. Here’s the thing: their new AI revenue now exceeds their SaaS revenue on a quarter-over-quarter basis. This is the roadmap.The Filevine Story: 10 Years of Grinding, Then AI Changed EverythingRyan Anderson didn’t set out to build a $3 billion legal tech company. He set out to stop waking up at 3am in a cold sweat.As a young trial lawyer in the early 2010s, Anderson was drowning. Deadlines piled up. Assignments disappeared. He’d lie awake convinced he’d missed something critical. “I’m not a naturally organized individual,” he’s said. “I’m naturally anxious.”So in 2014, he started building. First a Google spreadsheet — his “PI checklist” — at the law firm he’d founded with Nate Morris. Then a meeting over lunch in Las Vegas with Jim Blake, an engineer who asked the right questions: What’s breaking? Why is it so hard to keep track of work?That conversation became Filevine.For the next decade, they ground it out. Started with personal injury firms. Expanded into every legal practice area. Grew from task management to a full legal operating system: document management, demand generation, analytics, the whole lifecycle. By 2022, they’d raised $108M in a Series D — one of the largest legal tech investments ever at the time.Good company. Solid growth. But not a rocketship.Then AI happened.In September 2025, Filevine announced a $400M raise at a $3 billion valuation. The round was led by Insight Partners, Accel, and Ryan Smith’s Halo Fund. Smith — the Qualtrics billionaire and Utah Jazz owner — had been trying to invest for years. Anderson kept saying no. But after Filevine’s strongest quarter in company history, Smith called again: “You’re not getting your due.”What changed? AI revenue is now growing 130% year-over-year. Their AI chat product is growing 20%+ week over week. And as Anderson shared at SaaStr, their new AI revenue now exceeds their SaaS revenue on a quarter-over-quarter basis.Today: $200M+ ARR, growing 50-60%, 6,000 customers, 700 employees, 96% GRR, 124% NRR.This is what it looks like when a decade of building the system of record meets the AI moment.Top 5 Takeaways* “Sprinkling AI on top” is fundamentally wrong. You can’t just connect to OpenAI’s APIs and call it an AI product. That won’t cut it in 2026. You have to change your architecture.* Nothing is sacred. You will have to tear down meaningful components of working, revenue-generating code. Use the 4-quadrant framework: map every system against “competitive advantage” and “speed.”* Your SaaS is the closet, not the clothes. AI agents need context (your system of record), not just documents. This is your moat against AI-only competitors.* Protect your data and price to dominate. Move from open APIs to personal access tokens. Your high SaaS gross margins let you undercut AI-only competitors on blended margins. Be savage.* Obsess over usage, not revenue. No AI product goes beyond beta without audit trail logging. If customers aren’t using it, it doesn’t matter.The Wake-Up Call: “We Get to Sprinkle AI on Top”Ryan opened with a story that will sound familiar to many B2B and SaaS leaders:“I had an engineer say to me just a few months ago with a ton of pride, mind you: ‘We have built an incredible SaaS application that makes tons of money, grows fast, customers never leave it. We have almost 96% gross revenue retention, 124% net revenue retention.’ He has every reason to be prideful. And he said, ‘The great news is now we get to sprinkle AI on top.’”Ryan’s response? That is fundamentally incorrect.Connecting to OpenAI’s APIs isn’t going to cut it in 2026. To be AI-native, you have to change the architecture of your system. It has to flip.The Proof Is in the NumbersThe transformation is real and measurable. As Anderson put it at SaaStr AI London:“It is very plain to see that the numbers back up that we are now doing far more revenue on a new quarter-by-quarter basis in AI products than in our SaaS product. Now, that’s not to say that the SaaS product is in any way less successful — in fact it’s still growing at 35-40% year-over-year. We are just growing so much faster on the AI side of the house.”This isn’t a pivot away from SaaS. It’s SaaS + AI compounding together.Framework #1: The “Nothing Is Sacred” 4-Quadrant MatrixThe hardest part of going AI-native? Telling your teams that some of what they’ve built — things that work, that make money — has to be torn down.Ryan introduced a simple 2×2 matrix:Y-Axis: Critical to competitive advantage → Not critical to competitive advantageX-Axis: Keeps you moving fast → Slows you downThe Four Quadrants:Upper Right (Keep & Fortify): Critical to your moat AND keeps you fast. This is the cornerstone of your AI-native movement. Don’t tear it down — make it better.Bottom Left (Tear Down): Not critical to your moat AND slows you down. This is logically easy but emotionally brutal. These have to go.Upper Left & Bottom Right (Judgment Calls): More nuanced. Evaluate coldly, not based on feelings.The key insight: Someone who worked 5 years building a microservices architecture that doesn’t serve your AI needs will fight to keep it. You’ll have to be disagreeable as a CEO or technical leader making these calls.Framework #2: Content to Context (The Clueless Analogy)Here’s Ryan’s memorable way of explaining why “SaaS is dead” is wrong:“Imagine I came to you and said, ‘Hey, good news. We have an AI agent that can pick out your outfits in the morning.’ You’d be like, ‘Awesome. Done. Sign me up.’ In fact, Cher in Clueless already had this.But if you then said, ‘Oh, by the way, now that you have this AI agent, you don’t get to have your closet anymore. We’re not going to show you your closet. You can’t see it. It’s just a bunch of unstructured data and clothes and a mess.’You’d be like, ‘Hold on. I would actually like to have my closet AND the AI agent. Can I have both?’”Your SaaS application is Cher’s closet. The agent helps take action based on the content inside the closet.This is why SaaS companies have a significant advantage over AI-only competitors. You have:* The system of record* Audit trails (who did what)* User identity data* Deadlines and calendars* Contact information* Structured workflowsTo answer a simple question like “What should I do next on this case?” — you need ALL of this context. Documents alone give you an incomplete answer. And in most domains, an incomplete answer is actually worse than an inaccurate answer because the customer doesn’t know what they didn’t see.The Architectural Flip: AI Data LayerThe old architecture: AI layer sits on top of your core services, AWS, data, codebase, calendar, etc.The new architecture: AI Data Layer sits RIGHT NEXT TO the AI Application Layer.Why? Because your ML engineers need to tune and change how data flows into AI applications on nearly a daily basis. They can’t be going to your traditional tech team asking “Hey, can you please change how the API provides me this data?”The AI Data Layer owns:* How information is prepared for AI* How you ingest, process documents, emails, messages, and events* The graph structure of your domain (for legal: people, events, claims, outcomes)When ML teams own this data layer, Ryan says the results are “dramatically better, dramatically more reliable, higher context, more complete, more accurate.”This layer powers core AI applications:* Co-pilot* Search (semantic + traditional, without forcing users to choose)* Summarization* Recommendations* ReportingHiring AI Natives: The Data & Distribution PitchHere’s the problem: AI natives don’t want to work for “old SaaS companies.” They want to work for AI companies.Here’s the good news: The best AI natives actually want to work where they have more access to data.Ryan’s pitch to AI talent:* Data Access: “In legal tech, there are hundreds of competitors saying ‘give us your documents and we’ll run AI on them.’ But we know that to actually answer a legal question, you need way more than documents. You need audit trails, user identity, deadlines, calendars, conflict checks, contact information. We have ALL of that.”* Distribution: Show them what happens when you ship to an existing customer base. Filevine launched a product that went from 5-10 users/day to hundreds of users/day in just a few months. “Your AI team will love building products that absolutely rip because you have distribution and data.”The Acquisition OptionRyan’s confession: “We had an ML team. It was fledgling. Now we simply bought a company.”Filevine acquired Parrot, an AI-native company, and merged the teams. AI natives want to work next to other AI natives. Acquiring gives you critical mass fast.Rebrand With IntentFilevine changed their logo from a military/legal vibe to something bolder, “up and to the right.”But the real audience wasn’t customers — it was internal.“This change in our mark has told the people who work at Filevine every single day: the old mark is from a traditional SaaS era and the new one is from the AI era. It is highly symbolic. You should have no problem telling your team ‘we are moving’ — and you need to give them a symbolic thing to look at for that change.”They also created a new category: LOIS — Legal Operating Intelligence System. Not SaaS. Not AI. A blended category.Obsess Over Usage“We do not let our teams roll out applications beyond beta without audit trail logging to know exactly who’s doing what.”Filevine’s real numbers:* AI Fields product: 150 million actions in just a few months* Docker View product: Growing extremely quickly* Chat with your case (co-pilot): Their blockbuster product, growing 10% week over week in usageThe pattern they watch: A customer uses it 5 times, then 8 times the next day, then 20 times. Now they have customers using it 2,000 times a day.This is how you know if your AI product is any good: Are customers using it?Leverage Your Data: The API NegotiationAI-only competitors will come demanding your data “like it’s their moral right”:“I can’t believe you have all this data and you won’t give it to me for free whenever I want it. It’s your customer’s data. How could you possibly be acting this way?”Ryan’s response:* Control your APIs. Filevine moved from open API access to personal access tokens. They know exactly who’s accessing, what they’re doing, and how often.* Review every request. They’ve never said “no” to a competitor, but they always say “let’s have a conversation about that.”* Flip the script.“When they demand access, say ‘Okay sure. But of course it goes both ways, correct? We can take the AI outputs you get from our data and we’ll get them right back into our system. Correct? Isn’t that how it’ll work?’ All of a sudden, the shoe doesn’t feel so good when it’s on the other foot.”* Watch the API traffic. It reveals promising areas for new development. You’ll see which products are gaining traction. Copy those products and build them into your system. You have the right to do this.Price to DominateYour SaaS application likely has very high gross margins (Filevine: ~80%). Your AI-only competitors struggle with margins badly because of LLM costs.This means you can sell AI products at a lower price point than competitors.Why? Blended gross margins. Even if your AI gross margin drops to 30-40%, your blended margin might go from 80% to 60%. That’s still way better than an AI-only competitor whose margin is driven down to 10%.“Your investors might say ‘Why are we selling cheaper than AI-only competitors?’ Your answer is: because we’re gaining market share. And our blended gross is still higher than their blended gross.”Be savage on pricing. The AI-only competitors will cede you no ground.Build One Product, Sell to AI Customers OnlyThe boldest move Filevine made:“We no longer sell to customers who won’t buy the AI products.”Why?* Architecture simplicity: If you assume AI is implicit in everything you build, you don’t have to maintain two paths.* Team morale: How do you tell your SaaS engineers “You work on the old stuff while the ML team works on the cool AI stuff”? That doesn’t work. One company, one product.* Customer quality: “Show me the lawyer that doesn’t want to use AI and I will show you the lawyer that’s about to get his butt kicked.”They’re also moving from subscription/user-based pricing to usage-based pricing (per “matter” or project). More revenue from usage-based customers than traditional subscription customers.The 5 Biggest Mistakes SaaS Companies Make Going AI-Native* “Sprinkling AI on top” — Connecting to APIs without architectural change doesn’t make you AI-native. The AI data layer needs to sit next to the AI application layer, owned by ML engineers who can tune it daily.* Being too agreeable — You have to be disagreeable as a CEO when telling teams their working code has to go. Evaluate what to tear down logically and coldly, not based on somebody’s feelings.* Thinking documents are enough — AI-only competitors claim they just need your documents. Wrong. Documents alone give incomplete answers, and incomplete is worse than inaccurate because customers don’t know what they didn’t see.* Giving away API access freely — Move to personal access tokens. Review every request. Watch the traffic to see which AI products are gaining traction — then copy them and build them into your system.* Maintaining two products (SaaS + AI) — Build one product. Sell to customers who will come with you on the AI journey. If they won’t buy AI, they’re not worth selling to.The Bottom LineFilevine’s story is proof that SaaS companies can win the AI transition — but only if they treat it as a true transformation, not an enhancement.The companies that succeed will:* Tear down what doesn’t serve AI, even if it’s working* Shift from content systems to context systems* Flip their architecture so ML teams own the AI data layer* Acquire AI talent fast (even through M&A)* Obsess over usage, not just revenue* Protect their data advantage* Price aggressively using blended margins* Sell one integrated product to AI-ready customersAs Ryan put it: “At the end of the day, it has always just been about a customer with a problem. That’s what animates us. Can you solve my problem? We can solve it with technology.”The question isn’t whether you’re a SaaS company or an AI company anymore. It’s whether you can solve customer problems better than anyone else — using everything you’ve built plus everything AI enables.That’s the new game.Thanks for reading SaaStr AI: How To Sell, Scale, and Win! Subscribe for free to receive new posts and support my work. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

Philip Lacor is the CRO of Personio, a $3B+ HR and payroll platform with 1,500 employees, 15,000 customers, and a 400-person sales team. He shared their AI transformation journey at SaaStr AI London — and the learnings are a masterclass for any revenue leader trying to figure out how to actually deploy AI in GTM.We’re all hearing about AI-native companies crushing it. Replit, Gamma, Harvey.But what if you’re running a real B2B company? One with 400 salespeople, 15,000 customers, and years of accumulated process debt?That’s exactly where Personio was in May 2024 when their CEO kicked off an “AI Surge Week” — and what happened next is one of the most practical AI transformation stories I’ve heard.In just six months, they went from “90% of our team uses LLMs weekly” (which sounds good but isn’t transformation) to building 400+ AI assistants, cutting research time from 2 hours to 15 minutes per rep, and booking 140 meetings in 7 days through their AI SDR.Here’s what Philip learned — the stuff that actually worked, and the mistakes you should avoid.The 5 Lessons: What Actually WorksLesson #1: You Need Both Top-Down AND Bottom-Up MotionHere’s the trap most companies fall into: They give everyone access to ChatGPT, run some training, and call it an AI initiative.Personio did that too. Their AI Surge Week was a huge success — speakers from OpenAI, Mistral, AWS. Project teams building agents. Company buzzing with excitement.But then Philip noticed something: High usage isn’t the same as transformation.“After the AI Surge Week, we felt that although usage was high, this is maybe not enough to reach true transformation and to really fundamentally change the way we go to market.”The problem? Bottom-up motion alone can’t make the hard decisions:* Resource allocation — Who’s going to spend 40% of their time on AI initiatives?* Permission — Can people actually stop doing their old workflows?* Budget — Which tools do you actually buy vs. just test?* Prioritization — Of the 50 possible use cases, which 3 do you build first?This is why Philip started the “AI Powered Go-To-Market” working group in June — a top-down initiative to complement the bottom-up energy.The takeaway: Bottoms-up gets you experimentation. Top-down gets you scale. You need both.Lesson #2: Cross-Functional is Non-NegotiableThis one seems obvious but almost nobody does it right.Personio built a working group with three distinct capabilities:* Data & Systems Team — Owns infrastructure, Snowflake, the technical backbone* Revenue Operations + GTM Engineers — The bridge between tech and business (they have 2 dedicated GTM engineers now)* The Business — Marketing, Sales, Customer Success, the actual usersWhy does this matter? Philip saw both failure modes:“We have seen cases where our data systems team built things with LLMs but it was lacking the business context and therefore the models didn’t work very well.”And the reverse:“We had sales people who wanted to do something but originally they did not have the support from either data or systems or RevOps.”They deliberately made the working group large — 15 people — to get broad coverage across functions and build cultural buy-in.The takeaway: AI in GTM isn’t a sales project or a data project. It’s a cross-functional transformation. Build the team accordingly.Lesson #3: Use Jobs-To-Be-Done to Prioritize RuthlesslyHere’s what happened after they launched the Slack channel and started working on use cases:“People started to share opportunities, raising their hand, and the problem was that as people started to work on these new ideas, we hadn’t finished the first one. At one point it started to spiral a little bit out of control.”Sound familiar? Everyone gets excited, ideas flow, and suddenly you have 20 half-built things.Their solution: Jobs-to-be-done mapping.One of their GTM engineers literally shadowed account managers for two weeks. What she found:* AMs were working in 7-8 different systems to perform simple tasks* Constantly switching contexts, pulling information together* Losing 2.5 hours per day on one activity, 3 hours per week on anotherThey mapped every role’s jobs-to-be-done:* SDRs* AEs* Customer Success* Solution EngineersThen they overlaid these jobs onto the customer journey to see how they fit together — and where the biggest pain points were.The takeaway: Don’t just chase shiny AI use cases. Map your roles’ actual jobs, quantify the time waste, then prioritize based on where you have the biggest P&L challenge or customer experience gap.Lesson #4: Building an AI Culture Requires Leading, Sharing, and CelebratingPhilip has a formula he uses for transformation:Effect = Quality of Plan × Acceptance5 × 5 is way bigger than 10 × 1.So how do you build acceptance? Three things:Lead It: Philip does deal reviews where AEs used to show up with big PowerPoints. Now:“I would always go like, okay, please go to Gong, open up your account. There’s this little AI sign. Go in there. Now you look for the account brief and everything is there. And in real time, we would do away with PowerPoint.”The next time? Reps already know to use Gong’s AI. Leaders have to model the behavior.Share It: They put their own teams on stage to share what they’ve built:* An assistant to personalize customer decks* An assistant to answer RFPs* The expansion SDR assistantInternal success stories inspire more adoption than any training program.Celebrate It: This one’s clever: They announced that President’s Club will have 2-3 seats reserved for best AI contributions.Not sales performance. AI contribution. And next year? Even more seats.The takeaway: The #1 trait for an AI-powered GTM org? Curiosity. Hire for it, reward it, model it.Lesson #5: Great AI Comes From Your Stack + Your ContextHere’s an insight most people miss:“Let’s not go out and buy all these tools because usually the tools are not the panacea. There’s usually a lot of work that you need to do in your workflows, in your data.”Personio’s approach: Start with what you have, add LLMs on top, iterate from there.Their core stack:* Salesforce (or HubSpot)* Gong — They made a big bet here because “for a go-to-market organization, the customer conversation is obviously a very important source of data”* Qualified — Started with fast meeting booking, then layered on AI* Snowflake — Both structured and unstructured data* Amazon Bedrock — LLM layerBut here’s the work nobody wants to do:* One-third of their Salesforce data were duplicates — They installed automatic de-duping* Months cleaning the prospect database — Buying external data, connecting sources* Loading 5,000 Gong calls into Snowflake* Adding emails, connecting everything togetherThen — and this is critical — they added Personio-specific context:* ICP definitions* Pitch decks* Onboarding processes* Product training materials“This is really critical to really train the LLMs and to make it specific for your go-to-market, for your customers, and for your products.”The takeaway: The AI is only as good as the data and context you feed it. Clean your data. Connect your systems. Add your tribal knowledge. This is the work that makes AI actually useful.The 4 Use Cases That Actually WorkedUse Case #1: Win/Loss IntelligenceThe problem: Reps fill out Salesforce after winning or losing deals, but 30% of reasons were “Other” and even good data wasn’t deep enough.The solution: They loaded all conversation data, emails, and Salesforce data into Snowflake and built a GPT for go-to-market.The results:* Added 10-15% new insights to competitive battle cards* Created dynamic, continuously-updating battle cards (instead of static docs that go stale)* Data-driven product feedback: “Based on 10,000 calls, this is where we have weaknesses”This is evolving into a “go-to-market brain” — rep coaching, marketing campaigns, product prioritization, all from the same foundation.Use Case #2: Expansion SDR Assistant (2 Hours → 15 Minutes)The problem: Expansion SDRs were spending 2 hours per day researching customer information before making cross-sell calls. They’d check account health in Amplitude, contract details in another system, usage data somewhere else…The solution: A GTM engineer built an assistant embedded directly in Salesforce. Type in an account name, and it pulls data from 10-20 systems, formats it for cross-sell, and provides a recommendation (green/yellow/red).The results:* Research time: 2 hours/day → 15 minutes* Pipeline per FTE: ~2x increase* SDRs love it — “they’re using it every day and it makes the job better”Use Case #3: Intent Scoring for OutboundThe problem: Finding the right accounts, right people, right message is solved. But right time — knowing which prospect is actually in a buying cycle — is incredibly hard.The solution: Their data science team built a dynamic intent score based on multiple signals:* Website visits* Former users who moved to new companies* G2/Trustpilot activity* And other signals they continue to enrichThe score shows up directly in Salesforce with flame icons (🔥🔥🔥 = start here) and refreshes daily.Key learning: “We saw that initially the model was not great. It was picking up some signals that we didn’t think were good. We changed it and then it got better and better.”Use Case #4: AI Chat/SDR (“Nia”) — 140 Meetings in 7 DaysThe problem: Demo requests are your best leads. Waiting a week to book a meeting is crazy in a real-time world.The solution: They deployed Qualified’s AI chat (“Nia”) on their website. When prospects request demos, Nia books meetings immediately — 24/7.The results:* 140 meetings booked in 7 days* 200,000 website sessions processed* Deep insights into what customers actually ask about (pricing, product questions, etc.)But here’s what Philip found most valuable:“When you start reading the chats, you see all of a sudden customers have questions about your product. They want to know what your minimum price is. You’re getting very, very rich insights in what is top of mind for these customers. I got totally hooked on it.”The 24/7 reality: “People at 11 PM on a Friday evening, they’re thinking about requesting a demo. Why? But they do it.”The 5 Mistakes: What NOT To DoNow for the part that matters most — what went wrong and what to avoid.Mistake #1: Endless Tool Testing Without Going DeepThis came up multiple times:“I would not endlessly test tools. You got to dig in and go deep, learn from it.”Everyone wants to try Clay, then Artisan, then the next hot thing. But surface-level testing teaches you nothing. Pick a few tools and go really deep with them.“The point is not to test every single one of them. You got to pick a couple and go really deep with them. It’s all about the training. It’s all about the data.”Mistake #2: Learning AI vs. Doing AI“If you try to read all the papers and not do anything, I don’t think you’ll move fast.”The insights come from deploying, breaking things, and iterating — not from another podcast episode or Twitter thread.Mistake #3: Not Having Dedicated People Monitoring Agents DailyWhen they rolled out Nia, the AI chat:“There were definitely like maybe four weeks where we didn’t do enough and I don’t know how many demos we wasted by not training Nia really.”Now they have a dedicated person (“Ami”) who looks at output every day, applies feedback, tests in real time.Things they discovered only through daily monitoring:* Nia started giving legal advice (“Better not”)* Nia started bashing competitors (“That’s not us”)* When customers ask multiple things at once, Nia would answer the product question but forget to book the demo“You only learn that when you really start doing it and when you see where the AI stops.”Mistake #4: Building Without Business ContextTheir data systems team built LLM-powered tools, but:“It was lacking the business context and therefore the models didn’t work very well.”You can have perfect data infrastructure, but if the people building don’t understand the sales motion, ICP, or customer journey, the output won’t be useful.This is why the cross-functional working group matters — and why GTM engineers need both business and technical backgrounds.Mistake #5: Expecting AI Tools to Be Plug-and-Play“Usually the tools are not the panacea. There’s usually a lot of work that you need to do in your workflows, in your data.”Personio spent months:* De-duping Salesforce (1/3 of records were duplicates!)* Cleaning prospect databases* Loading conversation data into Snowflake* Adding company-specific contextThe tool vendors won’t tell you this. The AI is maybe 30% of the work. The other 70% is your data, your context, and your workflows.The Big Question: Can You Double Revenue Without Doubling Headcount?When asked about next year’s planning:“Managers say, ‘Hey, I need like 30 more people.’ That default should be, ‘Can I solve this with AI?’ The big question is: can we double the business with the same headcount?”That’s the real question every CRO should be asking.Philip’s honest take on where things stand:* Spending multiple six figures on AI tooling* Each SDR agent costs about $100K* Some teams will get smaller, others bigger (channel/partner teams can use more people)* “We will reallocate people” — it’s not about cutting heads, it’s about growing fasterA 6 Month Surge Is All It Took (To Really Get Going)Six months. That’s all it took for Personio to go from “AI Surge Week” to 400+ assistants, 2x pipeline per SDR, and AI booking 140 meetings a week.But here’s what actually made it work:* Top-down support for the hard decisions* Cross-functional team bridging data, systems, and business* Jobs-to-be-done to prioritize ruthlessly* Culture of AI — leading it, sharing it, celebrating it* Stack + context — not just tools, but your data and tribal knowledgeAnd what to avoid:* Testing tools endlessly without going deep* Learning AI instead of doing AI* Not monitoring agents daily with dedicated people* Building without business context* Expecting plug-and-play magicPhilip’s final advice:“The best career advice I can give you is lean into AI. What we do know is that everybody’s jobs will evolve, including mine.”The AI-native companies are moving fast. Your job, if you’re running a real SaaS company, is to move faster than they expect — and now there’s a playbook.Philip Lacor is the CRO of Personio. He flew from New York to London specifically to share this at SaaStr AI London, submitted through our AI speaker form (scored yellow the first time, had to resubmit to get green 🔥), and yes — the irony of an AI-powered GTM leader being evaluated by our AI speaker scorer was not lost on anyone. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

Jason Lemkin led the seed round via SaaStr Fund in unicorn Owner.com, an AI solution revolutionizing how small restaurants manage their business. Kyle Norton joined shortly thereafter, and after a slow few months, Kyle rocketed the org to almost $100m ARR in just a few years -- with growth accelerating at scale. Both Kyle and Jason have shared AI agents, learnings, and more on their AI agent journey and Kyle sat down with Jason on the very latest in AI for GTM. Kyle now manages a 100+ human AI-infused sales team and Jason and Amelia at SaaStr have deployed 20+ AI Agents.Top 10 Takeaways:* AI agents are now better than mid-pack AEs and SDRs. Not better than the best. But better than average. And that’s enough to fundamentally change how you build a GTM team.* The first agent is YOUR job. If you’re a CRO or CMO and you haven’t personally trained and deployed at least one AI agent, you will become obsolete. No agencies, no consultants. You. 30 days of work.* Pick one tool, not ten. The biggest mistake executives make is running 8-10 vendor bakeoffs. You can’t train 10 agents. Pick two—one incumbent, one startup—and go deep.* Salesforce is back—but not because of Agent Force. It’s because when you have 20 agents running autonomously, they need a hub. And Salesforce is that hub.* The middle is gone. You either work harder than ever to hit 10x5x5x5x growth rates, or you join a slow-growth company at 15-20%. The magical 2021 middle where you could have lifestyle AND exceed quota? That’s over.* Forward Deployed Engineers > Features. Don’t sign a contract until you’ve talked to the person who will actually deploy your agent. The best vendor isn’t the one with the best demo—it’s the one that will help you get into production.* Every agent takes 30 days to train. No shortcuts. You upload data, review outputs daily, correct mistakes, iterate. The agents that “don’t work” are the ones nobody trained.* Fix what breaks your heart first. Go to your website in incognito mode. Try to buy something. Try to get a question answered. Whatever breaks your heart—fix that with AI first.* AI-infused teams are 3x more productive. Kyle’s team at Owner is booking 3x revenue per AE compared to any team he’s ever managed. But that doesn’t mean fewer reps—it means higher quotas and more hiring.* The $250K SDR is coming. The elite folks—not the ones who think they’re elite on LinkedIn, but the ones who are genuinely 5-10x more productive—will earn 2-3x what they used to. But they’ll be expected to deliver 10x the output.The Backstory: Why SaaStr Went All-In on AgentsIt started with frustration.We had two salespeople making high market, six-figure salaries. They just quit going into our biggest event. No notice. No reasons. Just ghosted.I turned to Amelia, our Chief AI Officer, and said: “We’re done with this. I am done paying an SDR $150,000 a year or an AE $300,000 a year for basically inbound, spoonfed leads and renewals—and then having them quit on me.”Maybe you can be critical of me as a boss. Fair enough. But I’m pretty loyal. I pay people well. Do your job with me, and I’ll stick with you for 20 years. I just couldn’t do it one more time in my career.So we went all-in. Started in May with 1 AI Agent. Today we have 20+ agents running in production. They’re generating over $1 million in revenue. And here’s the scary part:Our AI agents are better than a mid-pack AE or SDR.Not better than the best. But better than the 50th percentile person I’ve worked with over my career. And that changes everything.The New Reality: Mid-Pack Sales Execs Are in Terminal DeclineLet me be blunt: if you’re a mid-pack GTM professional who doesn’t want to work harder and smarter than a year ago, these jobs are in terminal decline.We sent 70,000 hyper-personalized emails for SaaStr London using AI agents. They were better than the 7,000 emails humans sent before that. 10x the volume. Slightly better quality.And here’s what happened when we asked our highly-paid SDR to follow up on a lead I spotted on LinkedIn:“I’ll add it to my list and get to it when I can.”Half the time, they didn’t even follow up.The agent? The agent doesn’t argue. The agent just follows up.But We’re Still in Inning OneHere’s what most people don’t understand: what we’re doing today with AI GTM is just step one.Right now, “hyper-personalization” means maybe three dynamic fields in an email. One, really. Maybe we know your company name and your title.But imagine when AI really pulls in:* Every competitor you’ve ever used* Every page you’ve visited on our website for 10 years* Every interaction you’ve had with our brand* Every adjacent tool in your stackImagine when AI can send an email as good as the one that got me to invest in Owner—an email the founder probably spent several hours crafting with a top 0.1% IQ.AI should be able to do that. It’s just not there yet in GTM.When it is? Buy that product immediately.The Real Reason Agent Deployments FailThe failures of AI SDRs in 2024 were all LLM-based. The products literally didn’t work before Claude 4. It was slop.Now? They’re all above the line.So why do agents still fail today?Because people don’t roll up their sleeves and train them.We bought a RevOps tool. Didn’t train it. Didn’t pay attention. Thought it didn’t work. Then we got on a Zoom and I asked our highly-paid AE why we weren’t seeing any data.He said: “The app doesn’t work.”I said: “Do you see that Google link in the bottom left? You have to link up your account.”He linked it up. It showed he’d done nothing for 30 days. He quit that day.The tool worked fine. We just never put in the 30 days to train it.The 30-Day RuleEvery agent requires weeks of training before you can go live. Here’s what that actually looks like:Day 1-7: Ingestion* Upload your prospectus* Upload documentation* Connect to your database (or just your website as a base case)* The agent creates a list of questions—10, 20, 30 sample outputsDay 8-21: Iteration* Read every output. Every single one.* Correct what’s wrong* “Owner is great for 100-chain high-end restaurants” → Wrong. Fix it.* “Owner scales much more now, but our core audience is single-location restaurants with significant to-go business” → Correct.* The agent remembers. Every day it gets better.Day 22-30: Production* Hallucinations become a minor issue* You’re ready to scaleIf you don’t do this? You’ll say the agent doesn’t work. But the agent was never the problem.Do It Yourself, DudeHere’s my strongest advice for CROs and CMOs:If you don’t roll up your sleeves in the age of AI and AI GTM, you will become obsolete.This is not an agency game. Not today. Maybe in 24 months.I’m watching executives bring in their 11 agencies from their last job—their Salesforce agency, their outbound agency, their leftbound and rightbound agency from 2015—and they’re all failing.You are the agency. For now. At least for the first one.Here’s the brutal truth: if you haven’t trained an agent yourself, you have no idea what you’re talking about. Literally. You will be utterly ignorant in the age of AI.I did it. Even at this point in my career. I trained the first agent myself—every single day for a month. First thing in the morning. See what the agent said, see what’s wrong, start editing and changing.Then I proved it to Amelia. Then she did all the rest. She’s better and smarter than me. But I had to do the first one to even know what I was telling her to do.How to Pick Your First AI AgentStep 1: Pick one tool that solves a medium or higher-ranking problem.For many folks, it’s AI SDR. But it doesn’t have to be. Could be RevOps. Could be support. Pick something you’re passionate about—or pick whatever breaks your heart when you go through your own customer journey.Step 2: Find the right vendor partner.Don’t sign the contract until you talk to your Forward Deployed Engineer. I literally did this the other day: “I want to talk to my FDE. I don’t even need a demo. Who’s going to help me deploy this agent?”Still waiting to hear back. So that tool won’t be our first agent.If the vendor won’t connect you with your deployment team, find another vendor. I would rather have a worse vendor and know who’s deploying my agent than the world’s fanciest brand where I don’t know who’s going to do it.Step 3: Budget for $50-100K, not headcount.The first one is not about headcount. It’s about budget for one tool. You need 50K, 60K, or 100K—which is not nothing, but find the budget.Then you deploy it yourself, prove the ROI, and walk back to your CFO with data:“We just sent 70,000 automated emails. They’re better than humans. It generated 15% of the revenue for SaaStr London. Can I have some more budget?”Sure.The Two-Vendor BakeoffWe talk to so many folks doing bakeoffs of 8-10 vendors.In the old days, pre-AI, you could kind of do this. Buy SalesLoft, let the reps figure it out.Now? Each agent takes 30 days to train. You literally cannot do 10. It’s impossible.Do two.Pick one you already use (Intercom’s Finn, Zendesk’s AI, Salesforce’s Agent Force) and one hot startup that someone like you is using successfully.Ask for references. Send an email. Kyle from Owner will respond. Marshall from Manglement will respond. They’re all getting 50 reference requests a month. Just don’t ask for a call—send an email and they’ll give you a one-word honest answer.That’s enough.Because here’s the honest truth: all the leading agents are good. They’re all so much better than pre-AI that what matters is can you get it into production?These tools are evolving so quickly that during 2026 they’re going to converge. Feature parity used to take years. Now it takes weeks.So just pick the one that’s the right match for you. The one with the best deployment support. And go.The Lowest Hanging Fruit: Fix Your InboundIf you want the single easiest win, it’s adding AI to inbound.There is no excuse today for prospects not getting instant answers to their questions.There is no need for some 21-year-old SDR to qualify whether I’m worth their time to talk to an AE.Go to saastrannual.com. Talk to the digital Amelia—video, text, audio, however you want. She’ll answer your questions. If you want to sponsor, she’ll qualify you instantly. If you’re not a fit, she’ll tell you.No waiting. No scheduling calls. No hoops.I tried to buy a 10K product recently. Emailed the rep. They passed me to someone else. That person wouldn’t answer a single question until I got on a call.If they’d had an AI bot, I would have bought it in real time.That process is just not OK anymore. It’s no longer necessary. There’s no need to have high-friction sales that benefits some team’s funnel in theory.AI can score a lead better than a human. Today.Why Salesforce is BackKyle and I are both leaning into Salesforce more than ever. Here’s why: when you have 20 agents working autonomously, they need a hub. Somewhere for their data to meet. Somewhere to resolve conflicts.Salesforce is that hub.We use three different AI SDRs working with three different segments of our base. They all push their learnings back into Salesforce. They all share data. It’s a virtuous circle.The potential conflict? We pay more for those agents than we pay for Salesforce. Which is an existential question for Marc and his team. The agents are extracting the majority of the value.That’s why AgentForce has to win. That’s why they have 2,000 people working on it.But here’s my take after being one of the few in production with both Agent Force and competitors:* Agent Force is more work to set up* But it’s as good or better in production* The native integration with Salesforce data actually makes a difference* The emails are pretty similar across all the leading toolsIf Salesforce can make Agent Force more turnkey, it’s really going to win.The 3x Productivity QuestionKyle’s AI-infused team at Owner is 3x more productive on a per-AE basis than any team he’s ever managed.So does that mean hire fewer reps? Not exactly.If it were truly 3x, you could probably get away with a 30-40% smaller team and still hit your growth numbers. Not a third—but 30-40%. But the hottest AI companies can’t find enough people to hire. The quotas at OpenAI for the enterprise team are really high. And they still don’t have enough people.What’s changing is leverage.Traditionally in sales, there’s been no leverage. Every year you need to add more reps. In fact, there’s anti-leverage—it gets harder to get that incremental dollar. For folks that crack the code with AI, there may be leverage for the first time. And that means the elite folks—the genuinely 5-10x more productive ones—should be paid 2-3x more than before. No problem paying someone who manages 20 agents two or three times more than a traditional outbound director.But we’re expecting 10x the productivity. It’s not charity.Triple Triple Double Double Isn’t Enough Anymore. At Least Not for VCs.Let me be clear: triple triple double double is plenty good to build a multi-billion dollar company. Put it in a spreadsheet. 2→6→18→36→72. You’ll build something real.Three or four years ago, 50% of VCs would fund you at that growth rate.Today? Maybe 10%.It’s not 0%. But you have to find someone who really believes.The same amount of money is going into venture as the peak of 2021, but to less than half the companies. For startups, this means:* Whatever capital you have needs to last* Recruiting is harder because the best people want to work at the fastest-growing companies* You need to be ruthlessly honest about your fundabilityGo to saastr.ai/aivc. Upload your investor deck. It will tell you your exact percentage odds of getting funded based on all the benchmarks. A founder just below triple triple double double was confident they could raise. The AI said 38%. It changed their perspective.Pick Your PathHere’s the bifurcation that’s happening:Path 1: Work harder than you ever have.If you want venture-scale outcomes—10x 5x 5x 5x growth rates—be ready to work harder than ever. Kyle is working the hardest he ever has. I’m working the hardest I ever have. At Owner, even with all the tailwinds, even with an awesome team, Kyle’s still grinding.This is what it takes now.Path 2: Join something slow-growth.The folks growing 10-15-20% a year need people too. Good cash comp. The right amount of pressure. Taking a company from 20% to 25% growth is a big win for the year.If you don’t want to be thinking about work constantly, if you want to log off at 4 or 5pm—I’m not mocking that. We’re human beings. But pick your path.The middle is mostly gone.In 2020-2023, you could have it all. Lifestyle, work from home, 20-30 hours a week, exceed quota because of structural tailwinds. That doesn’t exist anymore.Even the fastest-growing startups are either lean, back in office, or high intensity. There is no magical middle.Be honest about what you want. Then pick.Top 5 Mistakes Execs Make with AI GTM Agents1. Running 8-10 vendor bakeoffs instead of picking two and going deep.You can’t train 10 agents. You don’t have 300 days. Pick your incumbent option and one hot startup, do real deployments of both, and decide.2. Outsourcing deployment to agencies or consultants.There are no AI GTM agencies that understand your business well enough to train your agents. Not yet. Maybe in 24 months. Today, you are the agency.3. Buying a tool and putting it on the shelf without 30 days of training.Every agent that “doesn’t work” is an agent nobody trained. You have to upload data, review outputs daily, correct mistakes, and iterate. Every single day until it works reliably.4. Not talking to the Forward Deployed Engineer before signing.The best vendor isn’t the one with the best features. It’s the one that will help you get into production. Don’t sign until you’ve talked to the person who will actually deploy your agent.5. Giving AI tools to individual reps to “figure out on their own.”A CMO at a $10B company told Amelia they were going to buy an AI SDR tool and just hand it to each SDR to figure out—no training, no centralization. That old paradigm of “buy SalesLoft, let reps run their own cadences” doesn’t work with agents. You need a nerdy GTM person at the top of the stack managing the whole thing.Quotable MomentsFrom Jason Lemkin“I am done paying an SDR $150,000 a year for basically reaching out to with mediocre emails to leads that are already high qualify—and then having them quit on me.”“Our AI agents are better than a mid-pack SDR, and in part, AE. Better than the 50th percentile people I’ve worked with over my career. And so we just don’t need them.”“If you don’t roll up your sleeves in the age of AI and AI GTM, you will become obsolete.”From Kyle Norton, CRO Owner“We’ve got nine high-impact production use cases of AI. Our booked revenue per dollar spent, on a per-AE basis, is 3x any team I’ve ever managed before.”“We built our cold outbound email program infrastructure in three weeks. Everybody else told us that was a multi-month project.”“Even at Owner, even with all the tailwinds, even with an awesome team—I’m still working the hardest I’ve ever worked.”Try our AI tools at saastr.ai — AIVC for fundability analysis, AI agents directory, and more. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

At SaaStr AI London, Amelia and I went deep on our AI SDR journey. We shared all our data, all the emails we’ve sent, all the performance metrics—everything. And the response was overwhelming.But here’s the thing: the #1 objection we kept hearing was “Yeah, but this won’t work for me. I don’t have your scale. I don’t have your data. I don’t have 10 years of history.”That’s simply not true.If you have customers, if you have revenue, if you have a database of any size—AI agents will work for you. You don’t need as much data as you think. You don’t need as much trailing history as you think. What you need is a methodology.Here’s what we’ve learned after sending 60,000+ hyper-personalized emails, booking 130+ meetings automatically, and generating 15% of our London event revenue through AI agents alone.The 5 Biggest Learnings From Deploying AI SDRs#1. AI Agents Crush the Work Humans Won’t DoThis is the single most important insight we’ve discovered.Our human SDRs wouldn’t follow up with return attendees for ticket sales. It wasn’t worth their time—they wanted to hunt six-figure sponsorships instead. We tried incentives. We tried Starbucks cards. We begged them. They said they’d do it, then we’d check the activity logs and discover they lied.The result? When we deployed AI agents on those exact same leads, they generated 15% of our London ticket revenue. Revenue we literally would not have gotten otherwise.Same story with our “ghosted” leads—people who reached out wanting to sponsor SaaStr for five and six figures, and our human team just... never responded. Not because they didn’t like the leads. Because every salesperson is force-ranking in their head, putting all their effort into the one big deal closing this quarter.The AI agent hit those ghosted leads with a 70% open rate.Here’s the mental model shift: Don’t think of AI SDRs as magic revenue generators. Think of them as the team that finally does the work your humans refuse to do. The small leads. The low-scored leads. The “not worth my time” leads. Those leads deserve better, and AI doesn’t discriminate.#2. Hyper-Personalization at Scale Actually Works—But “Pretty Good” Is Good EnoughBefore AI agents, our human SDRs sent maybe 75-300 personalized emails per rep per month. In six months with AI, we’ve sent nearly 60,000 hyper-personalized emails. That’s 32x the max human output.But here’s what people get wrong when they see our results: they expect jaw-dropping, month-of-research-level personalization.That’s not what this is.On a scale of 1-10, our AI emails are maybe a 3 to a 6 in customization. They’re pretty good. They reference the prospect’s company, what they’ve been looking at, maybe something they posted about. But they’re not poems. They’re not love letters.And that’s fine. Because the bar isn’t “better than the best human SDR having the best day.” The bar is:As good or better than your average human SDR, with 24/7 consistency.A lot of folks on the internet say “I could do better if I hired 30 top-tier Oxford graduates to craft one email each day.” Sure, maybe. But those people want to be promoted to AE in three months. They’re not going to stay. And you can’t hire 30 of them anyway.Pretty good emails with zero errors, sent consistently at scale, crushes inconsistent brilliance every time.#3. Train Your Agents Like You’d Train Your Best New HireHere’s where almost everyone fails with AI SDRs:They buy a product, do nothing, and expect millions in revenue.It didn’t work that way before Claude 4 when these products barely functioned. It didn’t work after Q1 2025 when they started getting good. It doesn’t work now.The way AI agents work for GTM is:* You figure out something that works with humans first* You nail the email, the script, the objections, the questions* You document what worked* You give it to the agent and train it for a month* Then you do it at scaleIf you’re expecting an agent to sell when you can’t sell, that’s never worked. Go back to founder-led sales basics. But instead of handing off to that first human hire, you hand off to your first agent hire.Same principles. Same rigor. Different execution.#4. Segment Ruthlessly—Never Unleash AI on Your Entire DatabaseThis is critical. Do NOT just point an AI SDR at your entire database and hit send.Here’s how we approach it:* Batch contacts into groups of 800-1,000 max for each campaign* Create sub-agents or sub-campaigns for each persona (CRO, CMO, website visitors, churned customers, etc.)* Train each sub-agent specifically for that persona and use case* Give each agent different goals (book a meeting, sell a ticket, follow up on a ghosted lead)Start with low-stakes segments:* People you ghosted* Good inbound you couldn’t fully follow up on* Post-meeting follow-ups that fell through the cracksDon’t start with mission-critical leads. You’ll be disappointed if you can’t get it working quickly, and these agents have ramp time.#5. You Need Exactly Two Humans to Make This WorkThis surprised us, but it’s become gospel:Human #1: A forward-deployed engineer from the vendor.Call them a solution architect, an FDE, whatever—you need someone from the vendor who will work with you on training and get your agent into production. If the vendor won’t give you this help, don’t buy from them. No matter how slick their sales pitch. A worse product with great implementation support beats a great product you can’t get working.Human #2: A GTM engineer on your team.This is the AI nerd. They could come from marketing (technical marketers, HubSpot nerds, anyone who’s built complex campaigns). They could come from RevOps if they’re technical enough. They probably can’t come from your standard sales team.Find the one GTM nerd on your team. Promote them. Have them own this. They’ll manage the orchestration—which contacts go to which agents, what CTAs, what follow-ups, what happens when leads close.Self-serve AI SDR products are coming, but we’re not there yet. Even Zendesk’s CEO told me their enterprise customers hit 60-80% automation with months of training, while self-serve gets 20%. Training with no humans isn’t quite ready.The Tech Stack That’s Actually WorkingWe run 20+ agents now. More agents than humans. Here’s the core:* Artisan: ~6% response rate on outbound* Qualified: ~6% response rate on inbound, 130+ meetings booked since August* Agentforce: 70% open rate on re-engagement (our newest agent, hitting ghosted leads)All of them required about two weeks to deploy and tune. All of them required ongoing spot-checking and training refinement. All of them are connected to a single source of truth so we know which agents get which contacts.On the chat vs. voice vs. video question everyone asks: Don’t overanalyze it. Our data shows about 85% prefer chat, 15% prefer voice. Chat is easiest to implement. Voice takes a bit more work (though we did our voice clone on 11 Labs in five minutes). Video is two orders of magnitude more work.Start with chat. Layer in voice when ready. Video might add trust for high-ASP sales. Just sequence them and stop debating.The Top 5 Mistakes We Made (And You Should Avoid)Mistake #1: We Kept Humans Too Long on Work They HatedFor six years, we tried to get human SDRs to reach out to return attendees about tickets. Incentives, begging, monitoring—nothing worked. They said they’d do it, they didn’t.The lesson: If your team consistently refuses to do certain work, stop fighting it. Deploy an AI agent on that segment immediately. That 15% of London revenue was found money we’d been leaving on the table for half a decade.Mistake #2: We Didn’t Read Every Message in the Early DaysWhen we first deployed, we assumed the AI would just work. We weren’t reading every single message our agents were sending.The lesson: In the first 30 days of any new agent, read everything. Every email, every chat response, every follow-up. You’ll catch errors, you’ll find training gaps, you’ll understand what’s actually being sent in your name. Only after you’ve built trust should you move to spot-checking and flag-based alerts.Mistake #3: We Underestimated Ramp TimeWe wanted instant results. The products promised quick wins. Reality was different.The lesson: Budget two weeks minimum to deploy each agent properly. If you get frustrated because “it should work in a day,” you’re setting yourself up for failure. This is training time that pays dividends for months.Mistake #4: We Almost Bet on People Who LeftGTM turnover was high before AI. It might be even higher now. We saw a CMO at a $50M ARR company who wanted to bake off 10 AI SDRs—and he was gone before implementation finished.The lesson: Don’t stake your entire AI go-to-market strategy on someone who might leave January 1st. If you’re building agents around someone (especially cloning their voice, training on their style), make damn sure they have a real stake in the company and a real reason to stay.Mistake #5: We Tried to Evaluate Too Many Vendors at OnceWe’ve talked to founders doing bake-offs with 8, 10, even more AI SDR vendors simultaneously. It’s chaos. Nothing gets properly trained. Nothing gets fair evaluation.The lesson: Pick three vendors max for any bake-off. Better yet, pick one that has a strong customer reference you trust, get the implementation help you need, and commit to making it work. Then expand from there.The Bottom LineIf you’re still having humans qualify prospects and waiting days for follow-ups in 2026, there’s no excuse. The products are good now. Chat, voice, even video—they all work.But this isn’t plug-and-play magic. It’s:* Take what humans have figured out* Document it* Train an agent with what works* Segment ruthlessly* Have two humans (vendor + internal) own the rollout* Read everything early, then build trust over timeEven if you grow just 15-20% faster in 2026 because of AI agents, it’s a gift from heaven. Because a lot of that growth is found revenue—the leads that weren’t being touched, the follow-ups that weren’t happening, the work humans just refused to do.Your leads deserve better. And now there’s no excuse not to give it to them.All our agent details, vendor breakdowns, and data are available at saastr.ai/agents. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

I did an open AMA at SaaStr London last week, a classic part of each SaaStr AI event. But this one was different. It was urgency to the max.The room was packed with founders, CROs, and marketers who all seemed to be wrestling with the same existential questions: Is inbound dead? Is the GTM playbook broken? Will AI agents replace my entire team? Should I just give up and become a forward deployed engineer?Most of the anxiety I’m seeing in the market right now is based on a false narrative. A dangerous “woe is me” narrative that’s been accelerating since late 2023. And I think it’s time to get honest about what’s actually happening—and what you need to do about it.The “Woe Is Me” Narrative Is Killing Your GrowthLet me start with the question everyone’s asking: “Is inbound dead? My traffic is down 50% in the last 12 months.”Here’s my honest response: Woe is you. Your SEO is harder. Woe is you. You don’t have as many leads as you had during a lockdown during a global pandemic. Poor you.This leads to a narrative that I think is quite dangerous: that the go-to-market playbook is broken and doesn’t work anymore.It’s just not true.Yes, the playbook that some folks are running from 2021 doesn’t work as well today. But here’s what I say: the plays all work. Webinars, inbound, outbound, leftbound, rightbound—it all still works.The Same CROs, CMOs, Etc. Are Running the Hottest AI CompaniesHere’s what’s fascinating: if you look at the hottest AI companies right now, you’ll see a cast of characters from the 2010s. B2B leaders you know from SaaStr 2017 and 2018 are running today’s AI rockets.* Vercel (just raised at $10B): Their COO? She was the Chief Business Officer at Stripe.* Replit (0 to $250M this year in vibe coding): Their CRO? He’s from ZoomInfo.* Bolt (one of the vibe coding leaders at $60M): Head of sales? He was on our old SaaStr sales team.That wouldn’t be possible if the plays don’t work. These leaders are using different tools. They’re using more AI things. But it’s the same playbook. Same demos. Same everything.The biggest real difference? There’s just so much demand. Tools like Cursor, Replit, Lovable, 11 Labs—they’re so disruptive that everyone is in market simultaneously. 11 Labs went from almost nothing to $300M this year. Bolt has so much inbound they can’t service it. At $60M ARR, Brian has maybe four people on his sales team. How many of thousands of leads can they follow up on?“They’re all classic B2B sales reps—just instead of calling every lead and trying to convince them their fungible product is the exact same as another product, they have insane demand and are servicing it. But it’s the same playbook.”The AI Budget Paradox: Record Spending, Record CutsHere’s the thing that can feel like a paradox but isn’t:According to Gartner, overall enterprise software is going to grow the fastest it ever has—15% a year at $400 billion. It’s never grown this fast ever. But of that 15%:* Almost half is taken up by price increases from existing vendors (everyone’s raising prices)* About half of the remaining half is new AI budgetThat means if you’re not one of the vendors getting price increases or new AI budget, everyone else has to get cut.Vendor count is getting stable or shrinking to make room for new AI offerings and price increases from select vendors. CIOs have gone around the room and said: “Give me an app. Give up an app. You want to add a couple AI apps next year? I’ll find you budget—but you got to give up two. You have 100 marketing tools? Maybe 96 is enough.”I just got an investor update yesterday from a pretty successful company at mid-eight figures in revenue. They had $1.5 million in churn last month. From happy customers. No CSAT issues, no other problems. They literally said: “We’re cutting apps next year. We’re an attachment to CRM and we got cut.”The CEO failed because they didn’t get above the cut line. It was great to have, but not mission critical. And it got cut to make room for an AI app.Our SEO Is Down 8%. But Our Traffic Is Up 50%.Is SEO dead? Let me give you our real numbers.At SaaStr, our blog is like our core—it’s home base. Our blog traffic was fairly flat for about four years. Going into this year, our SEO is down 8%. Not 50%, but 8% for real.But here’s the thing: our traffic is up 50%. We will end this year at saastr.com with twice as many readers as last year, even though our SEO is down.Why? Because people all want to read about AI GTM content. They don’t want to read about classic SaaStr themes about CS teams and CROs unless there’s an AI angle. But because we have some of the best AI GTM agent content out there, people are just devouring it.Meanwhile, G2 says their SEO is down 30-40% or more. So if that’s your only play and you haven’t changed your product or GTM since 2021, yes—it’s probably going to feel 8-50% harder.You’ve got to find your tailwind. That’s your job right now.AI SDRs Didn’t Work Until Claude 4. Now They Do.Someone asked me about AI SDRs: “You said they didn’t work, but now they do. Can you elaborate?”Really, none of these SDR products worked before Claude 4.1 Sonnet or Claude 4 this year. Replit didn’t work. Lovable didn’t work. Base44 didn’t work. They were out there—they just weren’t very good.Then the magical moment when Claude 4 came out and everything was kind of magical.Look at Gamma—they do AI presentations. Founded in 2020. It was five years to $1M, and then 1 to $80M this year. Replit was founded almost 10 years ago. It was 10 years to $1M, then $1 to $250M. It’s not a coincidence that all of these apps took off January, February, March of this year. That’s when the LLMs got better.If you had a bad experience with almost any AI LLM GTM product before March of this year, write it off. It was a different time. Different LLMs. Different world.The Two Failure Modes of AI AgentsI’ve identified two main failure modes for AI agent deployments:Failure Mode #1: Pre-Claude 4 LLMsIf you bought overhyped GTM SDR tools before February/March/April of this year, they just didn’t work well. The LLMs weren’t good enough.I was with the CEO of Qualified—they’ve been trying to use AI to qualify inbound leads since 2019. I asked him: “This really took off around February or March this year, right?” He said: “Yeah, that’s when it finally actually worked after 5 years.”Failure Mode #2: “Just Turn It On”The second failure mode: so many folks just told someone on their team to deploy it, or hired an agency that didn’t know what they’re doing.We were on a call with a global technology leader—pretty AI-forward, impressive company—and they said: “We’re thinking about buying our first AI SDR. We’re just going to buy it and hand it to our SDRs and have them figure it out.”It’s not going to work. You’ve got to train it. You’ve got to train it for a month. You’ve got to train it every day. You’ve got to iterate the onboarding.I’d say 80% of the conversations Amelia and I have are a version of this: “I bought a tool and I have no leads. I bought a tool and I have a lot of leads, but I didn’t connect it to my leads. It didn’t magically work.”The Right Way to Deploy AI AgentsHere’s the captain obvious thing for any AI SDR, BDR, even AI AEs, CSMs:* First, you’ve got to have it work in the real world with humans. Figure out what actually closes deals.* Then you tell the AI what worked. There’s no magical prompt. The prompt is: “Here’s the script that I use with a customer that I closed.”* Then you iterate that prompt.* Then you hook it up to data—Salesforce, HubSpot, Snowflake—and have it ingest the data to keep working with that prompt to make it better.* Read every email it sends. Read every response it ingests. Some will be dumb. Some will be wrong. You say: “Hey, that was wrong. SaaStr AI London is not December 4th.” Do it every day for about a month. The mistakes start to go away.“If you can’t sell it yourself, the AI can’t sell it for you. It’s that simple.”The #1 Skill for 2026: Become an Agent Deployment ExpertSomeone asked about the most important skill to build for 2026. Here’s my answer:Right now, if you are world-class at Agent Force or Clay or Qualified or Artisan—pick two or three leaders with decent revenue that people have heard of—and you go through the deployment yourself, you do the training, you onboard it, you get it working in your company, and you can point to the metrics…You will be infinitely hireable for the next 18 months because only like 2% of marketers have those skills.Become the expert in deploying agents. It doesn’t need to mean you know every agent or even needs to be technical, but you need to:* Pick a few leaders and deploy them yourself (not tell someone on your team to do it)* Watch it, iterate it, train it, iterate every day* Spend a week non-stop deploying it, then work on it every single day for a monthYou will learn so much and become infinitely deployable.The Benioff Insight: Deliver Insane Value Before They PayOn the first podcast that Mark Benioff did with us earlier this year, I asked him what he thought of Palantir. He said: “I’m jealous of how much they charge.” Pretty funny—they charge more than Salesforce does.But what he said next was even better: “I wish everyone at Salesforce could go live before they even pay.”Think about this. The way we used to buy B2B software for years: I need a CRM. Your friends use it. You saw it at your last company. You talk to a sales rep that could answer six questions. You’d buy it. And if you’re lucky, by the end of the year, you might deploy it.In the old days, you might have customers that were new for year two that never even deployed in year one. That doesn’t fly today.Mark’s point was: “I know I can’t deliver this today—it’s not feasible—but the technology is there. I wish every single customer would be live on Agent Force before they gave me a dollar.”Get as close to that today as you can for your customer.When you look at the companies that explode—Cursor, Gamma—think about how much value you get in the free program. Think about how much value you get in the first month. What can you do with AI to provide that much value to customers before they even pay you?If you provide insane value with an agent before they pay you, they’re going to be kind of happy to pay you—rather than the sucker bet we all made for two decades: “I think it works. Someone gave me a demo. The demo kind of works. I can’t use it myself, but I need that.” And then half the time it’s a horror story.Your Moats Are Shrinking From Years to MonthsSomeone asked: “Some AI experts say that with AGI in two-three years, SaaS will be dead because you can literally just talk to OpenAI and say ‘write code that does that.’ What do you think?”Listen, if you’d asked me six months ago, I’d say the odds that’s true are zero. That’s ChatGPT nonsense. You cannot reproduce the amount of workflows, corner cases, and complexity that complex B2B software does.But you can chip away at it.I’ve vibe-coded 11 apps in 90 days that have been used 800,000 times. I haven’t stolen revenue from anybody, but that’s a bit of an existential threat to somebody, isn’t it?And the agent has gotten so much better. When I built my first app on Replit 170 days ago, it blew up. It deleted my database. Those posts went viral on Reddit. We got death threats from Redditors. Millions of views.Now? When Replit launched V3 about 45 days ago, there’s multiple agents that talk to each other and check the work. Before it deleted my data, now it has an architect. When there’s a tough problem, it doesn’t just go off the rails—it says: “Hold on, I’m not sure. I’m going to call in another agent.” They bring in help on security. Now they have a design specialist.Don’t underestimate the rate at which this stuff is getting better.Competitive Edges Are Now Measured in MonthsWhen I started in SaaS, you’d build something slick and you’d have about 12-18 months until you were copied by a startup. Then you’d have like 5 years until a big company copied you.When we launched EchoSign, DocuSign (now an $18B company) only worked in Windows and was partially web-based. We had like 18 months before they decided to copy us. Then it took 5-6 years for Adobe to decide to copy it. Then Google just launched a clone last year—like a decade later.Not too recently, I invested in a startup I love. Incredibly powerful. Within two weeks, they had four clones. And then a massive company is building a clone that’ll be out this year.It’s not that we’re going to rebuild Workday and Salesforce and Oracle. But our competitive edges are going to be measured in months when they used to be years. And that compounds.Look at Google—they just launched a Replit/Lovable clone. Replit gets to $250M in one year, of course Google’s going to clone it. It used to take everything 5-7-10 years to get to $100M. No one would even bother to compete.“I don’t want to be one of these guys on X saying the only moat is speed and working 996. But there’s a lot of truth in it. If you don’t like it, you might go into terminal decay.”The Pager Duty Warning: Two Times Revenue at $500M ARRI wrote up PagerDuty today. It got about 500,000 views on X and LinkedIn.PagerDuty missed its last quarter. They just crossed $500M in ARR. They’re worth $1 billion. Two times revenue.Their customer count is flat—15,000-something customers for four years. DataDog came out with a clone. Atlassian’s product is better. Startups are better. They just didn’t move fast enough.And I think it’s only going to accelerate. Everyone will build clones faster. It took DataDog 15 years to decide to do this. It might take them 8 months to do it next time. And 90 days the time after that.Value Selling Is Dead Without Product ExpertiseSomeone asked about value selling: “Why is it so hard for sales teams to speak the language of CFOs?”Let me be honest in the age of AI: Most sales reps have no idea how to sell value.You cannot sell value unless you’re a product expert. Most sales folks want to talk with a war sheet, a tear sheet. They know six things. You cannot sell value if you’re not a product expert.We were with an AI leader the other day. They closed a seven-figure deal while we were there. The solution architect left the sales team behind. Closed it without them. Didn’t want to talk to the sales team. Sales team didn’t know the product. They added no value.Value-based selling means providing value. You cannot provide value in the age of AI if you do not know the product cold. Ideally, how to deploy it, how to get it going, how to deploy real value—not just having a valuation calculator on your website that says 18 months down the road the product will work.“The worst sales rep you can hire today is the one that tells you they’re a ‘great people person.’ Who cares? I want an AI SDR deployed in 30 days. I want it to get me this amount of quota. I want to do this workflow. The rep who says ‘Oh, it sounds good, yeah, we can do that’—it’s not going to work.”AI is not going to replace the AE the way it is already replacing the SDR, support, and customer success. But it is going to replace a lot of AEs that don’t know the product cold.I really think 70-80% of the sales executives I’ve worked with over the last five or six years don’t know their product cold. If you look back at your top couple sales reps—whatever era, pre-AI, doesn’t matter—the top ones weren’t just good schmoozers. They knew the product cold. I call them “sales magicians,” but they’re not magical. They just know how every nook and cranny works.We’re just not going to tolerate mediocre sales reps in the age of AI. We’re going to buy from somebody else.Sales, Marketing & Support Are Converging Into One AgentSomething fascinating is happening in e-commerce that’s relevant to everyone: In the space of just a year, three categories of B2B e-commerce have all combined—sales, marketing, and support.Think about it: when you go to a shopping cart, when you want to buy a new phone, what might you be doing? You might be buying (that’s sales). It might have a problem (that’s support). Or you might be interested but not ready today (that’s marketing).Those used to be different apps. In e-commerce, you had Klaviyo for marketing, Gorgias for support, separate apps for remarketing. About a year ago, you’d go to e-commerce sites and they’d have five different agents: “Talk here for support, talk here for marketing…”Now they’ve all combined because AI can easily understand: Is this a sales call? Is this marketing? Is this remarketing? Is this support? Is this customer success?In most B2B, we’re still running multiple apps. But that’s not going to last. People don’t want 28 agents on your website. They just want one agent to solve their problems.At SaaStr, we’ve gone from one to 20 AI agents ourselves this year. But it’s at the edge of too many—not only for us to process and manage business process change, but when you go to saastr.com or SaaStr Annual, we don’t want you to see 11 AI agents either.These discrete walls between categories—AI is going to break them down because we just want to talk with one agent as customers, prospects, and users.Outbound Still Works. Here’s Why.Someone asked about outbound: What’s changed? What stayed the same?From our AI outbound: geometrically more volume, same results. By training the AI with the best scripts and ideas that humans did, we basically saw the same results—in some cases a little worse, some cases a little better—but same as humans for the most part with much higher productivity.Outbound isn’t dead. And if you’ve never been a buyer at a big company, you don’t get outbound.Let me tell you a story. Back in the day at Dreamforce, a guy from Success Factors (a large SaaS company SAP bought) turned to me on a panel and said: “Yeah, I bought your product, Jason.” I said: “Wow, great. Why?” He said: “Well, it was the end of the quarter, our fax machines were broken, and you said you’d figured out e-signatures and I needed that problem solved today. I bought 300 seats.”One of his top problems at a given time. High ROI. Immediate value proposition. He read a cold email from a company at $2M ARR.I asked Yamini at SaaStr Annual this year: “Do you read your cold emails?” She said: “I read them every day. Email is the best. Even today—it’s the best thing in the world. Everyone reads it. It’s an open medium.”Your job with outbound is to solve one of the top three problems of your buyer.Whether it’s CEO, SVP, VP—if it’s a top three problem, you’re going to get a meeting. Whether it’s a top three pain point or a top three initiative. AI SDR, BDR, Agent Force, Artisan, Qualified—I don’t actually know if this is a top three problem for a lot of buyers, but I know it’s a top three thing on their mind. Everyone wants to figure out how to do an AI SDR.If you’re in the top three, you’re going to see super high open rates if you have a differentiated product and crystal clear value proposition. This has not changed in the age of AI.The Enterprise Budget Reality: VP Slush FundsLet me tell you how it really works in big companies. When I was a VP at Adobe, there were three types of budget: unbudgeted, discretionary, and budgeted.Budgeted money: Big dollars but really hard to get. Multi-million dollar contracts that often take 1-3-5 years of discussing and planning until you close the deal. Maybe get a pilot.Discretionary/Slush budget: I was the most junior VP of all 40-50 VPs. Even I had a $500K slush budget. Today it might be a million bucks. Everyone had more.What was my slush budget for? My top three needs that I just couldn’t get elsewhere in the organization or elsewhere budgeted through the CFO.If you solved one of my three needs sitting in my corner tower office and it was some fraction of that $500K-$1M—I had budget. I didn’t even care. It’s use it or lose it in big companies.I didn’t have one more dollar than that. And I didn’t care about my eighth problem. I just didn’t care. But selfishly, if you solved one of my problems, I had half a million to a million bucks to buy.Those are the budgets you’re fighting for. They haven’t changed in the age of AI, but our priorities have changed.It was really hard for me to get a million dollar piece of software—it probably would have taken me years as the new VP. But I had almost a million bucks in aggregate to spend on whatever I needed to get the hundreds of folks that worked for me to solve their problems. I could do that in a week or two. That was up until that line, and then I was tapped out for the year.The Bottom LineStop with the “woe is me” narrative. Stop saying the GTM playbook is broken. The plays all work. The same players from 2017-2018 are running the hottest AI companies—that wouldn’t be possible if the playbook was broken.What’s changed:* AI agents actually work now (post-Claude 4)* Moats are shrinking from years to months* You need to deliver value before you get paid* Product expertise is mandatory for sales* Multiple agent categories are converging into one* CIOs are cutting apps to make room for AIYour job right now is to find your tailwind. There is AI budget. There is AI budget for new vendors at the CIO level. If you have something that truly changes the game for efficiency—if you can truly replace lots of humans with your agent—your customers are going to want to meet with you and they’re going to want to buy.But if you’re running the exact same playbook as 2022, nothing’s changed, you’re an attachment that’s nice to have but not mission critical…You’re going to get cut.This post is based on the SaaStr London 2025 AMA. Join us at SaaStr AI Annual 2026 in May for more sessions on AI GTM, agent deployment, and the future of B2B SaaS. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

After deploying 5 AI SDRs across inbound, outbound, and follow-up—here’s the actual numbers, unexpected learnings, and what it really takes to make them workSix months ago, we had essentially zero AI SDRs at SaaStr. Today, we’re running five specialized AI agents that have sent nearly 20,000 outbound messages, closed over $1M in revenue, and fundamentally changed how we think about sales development.The results look incredible on paper: 6.7% outbound response rates (double the industry average), $1M+ closed in 90 days from our inbound agent alone, and 20% of our event ticket sales now coming from AI.But here’s what nobody tells you about AI SDRs: they require massive human oversight, they can’t fix what’s already broken, and the path to success is completely different than what vendors promise.SaaStr’s Chief AI Officer Amelia Lerutte and CEO Jason Lemkin share the real data, the brutal learnings, and exactly how we got these results. And want to see the tools we use? Click here.TLDR and Top 5 Learnings After Six Months of AI SDRs1. AI SDRs Scale What’s Already Working—They Can’t Fix What’s Broken* If your outbound isn’t working with humans, AI won’t save it* You must have proven messaging, defined ICP, and working processes before deploying* AI amplifies your best practices infinitely—but you need best practices first* We had to fix our broken RevOps processes before AI could help scale them2. They Require Massive Human Oversight (15-20 Hours Weekly)* These agents consume the signficiant amount of Amelia’s and Jason’s time to run successfully* Performance ebbs and flows directly with human attention—more time invested = better results* Weeks I’m busy with other work, agent performance noticeably dips* This is not set-and-forget technology; it’s coaching five SDRs simultaneously who work 24/73. Specialization Beats All-in-One. For Now.* We run 5 different AI SDRs, each trained for specific use cases (cold outbound, lapsed customers, active nurture, inbound qualification, ghosted lead recovery)* Even within one platform, we have sub-agents with completely different training* Specialized tools go deeper than all-in-one platforms—we’ll take three A+ tools over one B+ tool* The training specificity for each use case matters enormously for results4. The Unexpected Direct-Selling Capability* AI got surprisingly good at closing deals directly, not just booking meetings* For sub-$1K products (event tickets), our AI now closes deals autonomously* For higher ASP deals ($50-100K+), it qualifies and books meetings, then hands to humans* 20% of our event ticket revenue now comes from AI agents selling directly5. Budget $50-100K Per Platform + A Lot Of Your Time* Effective AI SDRs cost $50-100K+ annually per specialized platform* But the bigger investment is your time: 15-20 hours weekly managing them* We reallocated budget from two human SDR roles instead of finding new budget* ROI is clear (our inbound agent: $1M revenue in 90 days on ~$100K investment) but only if you commitThe Big Misconception Killing AI SDR DeploymentsThe myth: Buy an AI SDR for $50-100K, it magically generates leads, you replace human headcount, profit.The reality: AI SDRs scale what’s already working. They can’t create something from nothing.This is the #1 reason AI SDR deployments fail. Companies expect magic. They want to spend $20-100K and suddenly have leads pouring in without figuring out what messaging works, what audiences convert, or what their actual sales process should be.Here’s the truth that took us six months to fully internalize: Your AI SDR can only amplify your best practices. If your outbound didn’t work with humans, AI won’t save it.Think of AI SDRs as taking your A-tier sales development rep and giving them infinite time, perfect memory, and the ability to personalize at scale. But they still need to know what to say, who to target, and how your sales process works.We learned this the hard way. Before deploying AI, we had to:* Identify what outbound messaging actually converted* Clean up our RevOps processes (they were broken)* Define clear goals for each agent type* Create training based on real conversations that workedOnly then could we scale with AI. You can’t skip this step.Our 5 AI SDRs: The Specialized ApproachMost companies think about “an AI SDR.” We run five, each specialized for different use cases:Agent #1: Outbound Cold (Artisan)* Pure cold outreach to new prospects* Highly personalized based on company signals* Goal: Book qualified meetingsAgent #2: Lapsed Customer Outreach (Artisan)* Targets previous sponsors/attendees who haven’t engaged recently* Leverages past relationship for warmth* Goal: Re-engage and convert to new eventsAgent #3: Active Nurture (Artisan)* Follows up with people opening emails but not converting* Tracks engagement signals* Goal: Move them from awareness to actionAgent #4: Inbound Qualification (Qualified)* Lives on our website, engages visitors in real-time* Qualifies intent, books meetings, sells tickets directly* Goal: Convert inbound interest instantlyAgent #5: Ghosted Lead Recovery (Salesforce Agent Force)* Follows up with leads our human team dropped* Leverages full Salesforce history for context* Goal: Recover lost opportunitiesEach agent is trained completely differently. Different messaging, different collateral, different success metrics. This specialization is why they work.The Outbound AI SDR: Real Numbers After 6 monthsOur outbound agents (primarily Artisan) have now sent nearly 20,000 messages. Here’s what actually happened:Core Performance Metrics:* 19,847 total messages sent in six months* 6.7% overall response rate (industry average ~3%)* 4% positive response rate (significantly above platform benchmarks)* 3,000 emails per month from AI vs. 75-285 per month from previous human reps* 10% of London ticket revenue attributed to outbound AI aloneWhat This Actually Means:Our AI SDR sends more emails in one month than our best human SDR sent in 40+ months. And it does it with better response rates.But here’s the critical nuance: These results required massive human input.On weeks when I spend more time training the agent, reviewing outputs, and feeding it better contact lists—performance jumps noticeably. On weeks when I’m slammed with other work (like preparing for SaaStr London), performance dips.The AI isn’t truly autonomous. It’s a force multiplier for human expertise.The 5 Sub-Agents Strategy:Even within “outbound,” we don’t run one generic agent. We run five specialized versions:* Lapsed Sponsors: “We worked together on SaaStr Annual 2023, here’s what’s new...”* Current Sponsors: “You’re sponsoring Annual, have you considered London?”* Previous Attendees: “You attended last year, early access to this year...”* Engaged Non-Converters: “You’ve opened our last 5 emails about speaking...”* Pure Cold: “You’re building [specific product], here’s why SaaStr...”Each has different training, different tone, different proof points. The specialization matters enormously.The Unexpected Learning: Direct SellingWe initially deployed our outbound AI to book meetings for sponsorships. That worked fine.But then something unexpected happened: For lower-priced products (event tickets under $1,000), the AI got really good at closing deals directly.At first, I was nervous. “Can I trust the AI to sell without human review?” Six months in, the answer is yes. For sub-$1K tickets, it closes deals on its own. I let it run free now.For higher ASP deals (sponsorships $50-100K+), it still books meetings and qualifies, then hands to humans. But the direct-selling capability on lower-ticket items has been game-changing.The Deliverability Secret:One critical learning: Artisan forces a 2-3 week warm-up period before sending at volume. This annoyed me initially. “Why am I waiting 2-3 weeks? Just let me send!”Now I understand. Our emails hit primary inbox, not promotions tabs. Our deliverability is essentially perfect. I can’t even achieve this level with Marketo.Skip the warm-up at your peril. Deliverability is everything in outbound. The best message in the world doesn’t matter if it hits spam.How We Feed the Beast:The #1 operational challenge: Constantly feeding the AI fresh, quality contacts.I started uploading contact lists once per week. Now I do it twice per week when possible because the AI performs better with fresh inputs.About 90% of contacts are ours—from our database, not scraped from Apollo or other intent data providers. We trust our data quality, and that trust shows in results.I upload via CSV in batches of 800-1,000 contacts. This size seems to be the sweet spot for performance in Artisan specifically.The Inbound AI SDR: The $1M SurpriseWe added our inbound AI SDR (Qualified) in August—three months after starting outbound. This agent has produced the most surprising results of any deployment.The Numbers (August-November, 3.5 months):* 697,000+ sessions with website visitors* 1,000+ meaningful conversations (vs. simple questions)* 100+ meetings booked (approaching 100 as of November)* $1M+ in closed revenue in just 90 days* $2.5M+ in pipeline attributed to agent-booked meetings* 70% of October’s closed-won deals came from this AI agentRead that last stat again: In October, 70% of our closed revenue came through our AI SDR.Why This Agent Crushes It:The inbound AI SDR isn’t just booking meetings faster (though it does—instantly vs. up to 24 hours delay previously). It’s completely transforming the quality of those meetings.Here’s what changed:Before AI (The Old Way):* Prospect fills out contact form* Goes into queue for me to round-robin to rep (delay: minutes to hours)* Rep gets assignment, responds (delay: hours to 24 hours)* Back-and-forth to find meeting time (delay: days)* Meeting finally happens, first 10 minutes wasted on basic discovery* Deal cycle startsTotal time to meeting: 1-3 days Discovery needed: 10+ minutes Context provided: MinimalAfter AI (The New Way):* Prospect visits website, AI engages instantly* AI qualifies, understands needs, books meeting—all in real-time* AI provides complete dossier to sales team before meetingTotal time to meeting: Seconds to minutes Discovery needed: Zero Context provided: EverythingThe Context That Changes Everything:Before each meeting, we now know:* Complete conversation history (what they asked the AI, what they cared about)* Every page they visited on our website* How many times they’ve engaged over what timeframe* Other people from their company who visited (CEO browsing speaking opportunities while CMO books meeting)* Specific content they consumed (sponsorship packages, speaker guidelines, ticket options)We don’t do discovery calls anymore. The AI did discovery. We jump straight to solution discussions.Real Example:Prospect books a sponsorship call. Before the meeting, the AI tells us: “Their CEO was also on the site yesterday looking at speaking opportunities, even though this person only asked about basic sponsorship.”We mention this in the call: “Hey, we noticed your CEO was also checking out speaking. Should we look at a package that includes a speaking slot?”Prospect: “Oh wow, I didn’t know they were looking at that. Yes, let’s include speaking.”Instant upsell. This happens constantly now.The Training That Makes It Work:Most companies deploy Qualified with two buttons: “Talk to Support” or “Talk to Sales.” That’s it. Their agents are mediocre.Our AI ingests everything:* 20 million words across SaaStr.com, SaaStr.ai, London, Annual sites* Our entire YouTube channel* Recorded sales calls I upload* Sponsor meetings transcripts* Custom documentation and FAQs* Historical email threadsThe agent is empowered to:* Sell event tickets directly (up to $1,000)* Offer discount codes* Follow up if codes aren’t used* Book meetings for sponsorships* Route to support when appropriate* Remember returning visitors and full contextThe Discount Code Workflow:This was an unexpected use case that emerged from the data.Week one with the inbound AI, I noticed the #1 question was: “Can I get a discount on tickets?”I empowered the agent to give discounts and sell directly. Here’s what happens now:* Prospect asks for discount* AI: “Absolutely! Here’s code LONDON25 for 15% off. You can use it at checkout.”* Prospect leaves without buying* Day 3: AI follows up: “Hey Jason, I gave you code LONDON25 when we chatted. I noticed you haven’t used it yet. Still interested in attending?”* Prospect convertsThis follow-up conversion happens at scale. The AI remembers every interaction, knows who didn’t convert, and follows up systematically.Result: 20% of our London ticket revenue comes from AI agents (both inbound and outbound combined).I physically could not do this level of personalized follow-up at scale. The AI does it effortlessly for thousands of prospects.The Follow-Up AI SDR: Recovering Ghosted LeadsOur most recent deployment (October) was Salesforce Agent Force for a use case I’m embarrassed to admit we needed: following up with 1,000 leads our human team completely ghosted.The Embarrassing Reality:After SaaStr Annual, I audited our Salesforce. I found ~1,000 people who:* Filled out our “I’m interested in sponsoring” form* Were automatically routed to a sales rep* Never received any human follow-up whatsoeverThis is common. It’s also inexcusable. These were warm, inbound, high-intent leads. We just... forgot about them.The Agent Force Solution:These people were already in Salesforce with full interaction history. Perfect use case for Agent Force because it knows everything Salesforce knows.Early Results (One Month Live):* 72% open rate (unheard of in Marketo or cold email)* Higher response rate than our other agents* Still working through the initial 1,000 at a controlled paceWhy such high open rates? Because Agent Force personalizes based on complete Salesforce history:* Past event attendance* Previous sponsorship levels* Interactions with our team* Account tier and company information* Engagement patternsThe emails don’t feel like “recovered leads” outreach. They feel like natural continuation of relationship.Sample Email:Hi Kyle,I noticed you reached out after SaaStr Annual about sponsorship opportunities, but somehow we never connected (entirely my fault!).I see you attended Annual 2022 and 2023—thanks for being such a consistent supporter. Based on your company’s growth since then, I think our London event in March might be perfect timing.Would you be open to a quick call about 2025 opportunities? Here’s my calendar: [link]Best, Amelia (via AI)Simple, personal, acknowledges the gap, moves forward. Response rate is significantly higher than cold outreach.The Setup Reality:People think Agent Force is too technical. “You need a Salesforce admin.” “It’s for enterprises only.” “Setup takes months.”I’m not a Salesforce admin. I’m not certified. I was better at Marketo than Salesforce before this project.With Salesforce’s team help during onboarding, we got it working in days, not months. The key: I copied our Artisan training instructions, adapted them for “ghosted lead recovery,” and it worked immediately.What It Actually Takes: The Human CommitmentHere’s the part most AI SDR vendors gloss over: These agents consume the majority of both mine and Jason’s time.Could we run more agents? Yes. Would they fail without our oversight? Absolutely.The Weekly Time Commitment:For me personally, across all five AI SDRs:Outbound Agents (Artisan):* 3-4 hours weekly uploading and preparing contact lists* 2-3 hours reviewing performance, adjusting training* 1-2 hours reviewing draft responses for high-value prospects* 30-60 minutes monitoring responses and routing to humansInbound Agent (Qualified):* 1-2 hours weekly reviewing conversations, identifying gaps* 1 hour uploading new training materials (calls, docs, FAQs)* 30 minutes spot-checking responses* As-needed monitoring when agent raises hand for helpFollow-Up Agent (Agent Force):* 1-2 hours weekly preparing and uploading contact segments* 1 hour reviewing performance and adjusting targeting* 30 minutes monitoring send patternsTotal: 15-20 hours per week actively managing five AI SDRs.This is why we can’t just infinitely add more agents. The human oversight is real and necessary.Performance Ebbs and Flows with Human Input:We have clear data on this now: Agent performance directly correlates with human attention.Weeks I spend more time:* Response rates increase 10-20%* More meetings booked* Higher quality conversations* Better revenue outcomesWeeks I’m slammed with other work:* Agents still run (that’s the beauty)* But response rates dip* Fewer meetings convert* Revenue impact decreasesThey don’t fail catastrophically without me. They just perform at B+ level instead of A+ level.The Training Never Stops:Every week, I’m:* Adding new proof points that worked in human conversations* Removing messaging that got negative feedback* Updating targeting based on what’s converting* Adding new use cases and capabilities* Refining objection handling based on real responsesThis isn’t set-and-forget technology. It’s like having five SDRs who need constant coaching—except they never complain, never quit, and work 24/7 once trained.The Specialized vs. Generalist DebateA common question: “Should I get one all-in-one AI SDR tool or multiple specialized ones?”Our philosophy: Specialized over all-in-one.Yes, it’s more work. Yes, I have to manage separate tools and avoid contact overlap manually. Yes, it’s sometimes annoying.But specialized tools go deeper. Artisan is maniacally focused on outbound results. Qualified is obsessed with inbound conversion. Agent Force leverages Salesforce data better than any third party could.The Tradeoff:All-in-one platforms promise simplicity. “One platform for all your AI SDR needs!”In practice, we’ve found that platforms trying to do everything do each thing at B+ level. Specialized platforms do one thing at A+ level.For our use cases and our scale, we’ll take three A+ tools over one B+ tool. But this requires accepting operational complexity.The Contact Overlap Problem:One major challenge with multiple agents: Making sure the same person doesn’t get hit by three different AI SDRs.Right now, this is mostly manual. I carefully segment:* These contacts go to Artisan* These contacts go to Qualified* These contacts go to Agent ForceEach platform has de-duping within itself. But across platforms, I’m the de-duping layer.This is improving. Artisan just added the ability to exclude specific Salesforce campaigns. Qualified syncs natively with Salesforce. They’re moving toward better interoperability.But today, if you run multiple specialized AI SDRs, expect manual coordination work.The Budget Reality: What It Actually CostsReal talk: Effective AI SDRs cost $50-100K+ per platform annually.Breakdown typically looks like:* $60-70K annual subscription* $20-30K training and onboarding* Or ~$100K all-in depending on vendorSome vendors are launching cheaper self-serve versions. We’ll test these in 2025. Initial hypothesis: They’ll work reasonably well for simple use cases but won’t match enterprise power because they ingest less data and require less customization.Think of it like Zendesk support agents. The $299/month version works but only has 20% of the enterprise capability because it ingests your wiki vs. a decade of customer interactions and call transcripts.How We Funded This:We didn’t get new budget. We reallocated existing budget.Specifically: When SDRs left naturally after our Annual event, instead of replacing that headcount, we invested in AI SDR platforms.We replaced the budget for two human SDRs with our AI SDR stack. The AI sends 10x more messages with better response rates, so the math works.The ROI Calculation:Is $100K for an AI SDR worth it?Let’s do the math on our inbound agent:* $100K annual cost (rough estimate)* $1M closed revenue in 90 days* 10X ROI in one quarterEven if you cut those results in half—even if they’re 75% lower—the ROI is clear.But you have to commit to making it work. You can’t deploy and forget.The Vendor Selection FrameworkCritical advice: Don’t tolerate mediocre sales reps in the AI age.The best AI companies have shockingly bad sales teams that don’t understand their own products. Sales reps will tell you things that are flat wrong about capabilities, training requirements, or integration.What to Demand:* Talk to the actual implementation specialist or technical person before signing—not just the sales rep* They should assess your data and confirm success viability in 20 minutes: “Yes, you have enough data for this to work” or “No, you need X first”* Ask them how many deployments they’ve done personally and what success rate looks like* If a sales rep blocks access to technical experts, walk away immediatelyWe passed on one excellent AI SDR vendor because their sales rep was incompetent, didn’t understand the product, and created barriers to talking with technical teams.That rep cost their company all the PR, revenue, and referrals we would have driven. Don’t reward bad sales behavior with your budget.The Deployment Partnership:Every vendor we use provided hands-on setup help:* Artisan connected us with their implementation specialist* Qualified did the same* Salesforce/Agent Force provided onboarding resourcesThis is standard and necessary. The vendor should be invested in your success and provide technical resources to ensure it.You’re not figuring this out alone. Good vendors know this and staff accordingly.The “Too Much Demand” Problem:Interesting dynamic: All these vendors have more demand than they can handle.Some turn away business even when you have budget. Main reasons:* Not enough data to train effectively* Use case doesn’t fit their platform well* They’re at capacity and prioritizing customers most likely to succeedThis is actually good. It means they care about success rates more than just revenue. But it can be frustrating if you’re turned away.If a vendor says they can’t support you, ask why and listen. They’re usually right about whether you’re ready.What Actually Works: The Implementation PlaybookAfter six months and five AI SDR deployments, here’s the playbook that works:Step 1: Identify What’s Already WorkingDon’t deploy AI to fix broken processes. Deploy it to scale working processes.* What messaging gets responses from humans today?* What audiences convert at acceptable rates?* What does your best SDR do that works?* What sales process actually closes deals?Document this. This becomes your AI training foundation.Step 2: Start With One Agent, One Use CaseDon’t try to deploy across inbound, outbound, and follow-up simultaneously. Pick one:* Pure outbound if you have contact lists and proven messaging* Inbound if you have website traffic and can define qualification* Follow-up if you have a database of unconverted leadsGet one working phenomenally before adding a second.Step 3: Choose 1-2 Vendors MaximumDo a bake-off if needed, but limit it to two vendors you’ll properly train and compare.We talked to a CMO doing 10 simultaneous vendor trials. That’s insane. You won’t train any of them properly. The bake-off will fail and you’ll conclude “AI doesn’t work.”Two vendors maximum. Train them properly. Make an informed decision.Step 4: Take Your Best Person and Learn TogetherDon’t hire a “Chief AI Officer” initially. Don’t delegate to someone who doesn’t understand the work.Take your best SDR, sit down together, and figure out AI together. Learn by doing.Eventually, that person’s role will evolve to focus more on AI operations. But start as partners.Step 5: Commit to 90 Days of Daily ManagementPlan for this to consume significant time for three months:* Daily monitoring of outputs* Weekly training updates* Constant refinement of messaging and targeting* Regular review of conversations and responsesThis is not set-and-forget. It’s coaching five SDRs simultaneously.Step 6: Empower GraduallyStart with AI in draft mode:* It suggests messages, you approve and send* You review every interaction* You correct and train constantlyAfter 30-60 days of this, start empowering:* Let it send to certain segments without approval* Let it handle objections independently* Let it close small deals directlyWe’re six months in and still have some agents in draft mode for high-value prospects while others run autonomously for lower-stakes interactions.Step 7: Scale What WorksOnce you have one agent crushing it, add a second with a different use case.We went:* Outbound first (May)* Inbound second (August)* Follow-up third (October)* Now adding more use cases within existing platformsEach one took 60-90 days to reach peak performance. Don’t rush this.The Mistakes That Kill AI SDR DeploymentsAfter watching dozens of companies try and fail with AI SDRs, here are the fatal mistakes:Mistake #1: Expecting Magic Without Work“I’ll buy this AI SDR, it’ll generate leads, I’ll make money.”No. You’ll buy this AI SDR, spend 20 hours per week training it, constantly refine it, and then it’ll generate leads.The companies succeeding with AI SDRs are putting in massive human effort. The companies failing expected automation without investment.Mistake #2: Deploying to Fix What’s BrokenIf your outbound doesn’t work with humans, AI won’t fix it.If your messaging is off, your ICP is wrong, your offer is weak—AI will just scale your failure.Fix the fundamentals first. Then scale with AI.Mistake #3: Generic Training“Here’s our website, here are some email templates, go!”That produces mediocre results.Winning training:* Specific proof points from real conversations that worked* Objection handling based on actual objections you’ve received* Clear escalation rules for when to loop in humans* Detailed ICP definition with examples and non-examples* Response frameworks that match your brand voice exactlyMistake #4: Set and Forget“I deployed it three months ago and it’s not working.”When did you last update the training? What have you refined based on results? How often do you review conversations?“Uh... I deployed it and haven’t touched it since.”That’s why it’s not working.Mistake #5: Ignoring the Vendor’s ExpertiseEvery vendor knows things about their platform you don’t. They’ve seen hundreds of deployments.When they say “You need a 2-3 week warm-up period,” don’t ignore it. When they say “This feature won’t work for your use case,” believe them. When they suggest a specific training approach, try it their way first.You can innovate later. Start by following their proven playbook.The Future: What’s Coming NextAgent-to-agent communication is the next frontier.Right now, our five AI SDRs don’t talk to each other. I manually prevent overlap. This is improving but still mostly manual.Within 6-12 months, I expect platforms will communicate better:* “This prospect is already in an Artisan sequence, don’t add to Qualified outreach”* “This person just had a positive inbound conversation, suppress outbound”* “This account is in active deal cycle, route all touches to assigned AE”The infrastructure for this exists. The integrations are coming.Voice and video agents are next.We’re filming with Qualified to turn our chat agent into a full video/voice agent. It’ll have my voice, my face, and conduct two-way conversations.This should roll out by SaaStr London in March. Come see it in person.Lower-priced self-serve versions from every vendor.The $100K enterprise version will stay for complex use cases. But $299-999/month self-serve versions are launching across the board.We’ll test these in 2025 and report back. They won’t match enterprise capability but might be good enough for smaller teams or simpler use cases.The consolidation question.Will we eventually move to one platform that does everything? Maybe, if one gets good enough at everything.Right now, specialized wins. But I could see a world where one platform nails inbound, outbound, and follow-up at A+ level and we consolidate.That’s probably 12-24 months away.The Honest Assessment: Is It Worth It?Unequivocally yes, but only if you commit.Our results after six months:* 20K+ messages sent vs. * $1M+ closed revenue from inbound agent in 90 days* 10X scale on activities that were already working* Better conversion rates than human-only in many cases* 20% of ticket revenue from AI agentsBut this required:* 15-20 hours weekly from me managing the agents* Deep commitment to training and iteration* Willingness to trust agents with real revenue operations* Tolerance for failure and public criticism* Significant budget reallocation ($200-300K+ across platforms)* Six months of continuous learning and improvementThe magic isn’t that AI SDRs work without effort. The magic is that once you invest the effort to train them properly, they scale your best practices infinitely.The ROI is clear if you do the work. Our inbound agent alone generated 10X ROI in one quarter. Even cutting that in half or by 75%, the math works overwhelmingly.But there’s no shortcut. You can’t buy an AI SDR and expect it to magically work. You have to train it like you’d train your best human SDR—except this one never sleeps, never forgets, and gets better every day.Your Next StepsIf you’re considering AI SDRs:Week 1: Assess readiness* Do you have something working that needs scale? (If no, stop here)* Do you have data to train on? (6+ months minimum)* Can you commit 10-20 hours weekly for 90 days? (If no, wait)* Do you have $50-100K budget? (If no, wait for self-serve versions)Week 2-3: Vendor selection* Narrow to 1-2 vendors based on your primary use case* Talk to technical teams, not just sales reps* Get them to assess your data and confirm viability* Check references from similar companiesWeek 4: Start with one use case* Outbound if you have proven messaging and contact lists* Inbound if you have traffic and clear qualification criteria* Follow-up if you have unconverted lead databaseMonths 2-3: Train and iterate daily* Review every conversation initially* Refine training based on real results* Add proof points from what works* Remove messaging that failsMonth 4: Start empowering* Let it run autonomously for low-stakes interactions* Keep human oversight for high-value prospects* Measure results vs. benchmarksMonth 5-6: Scale or add second use case* If first agent is crushing it, add a second* If first agent is struggling, go deeper before expanding* Never deploy more agents than you can actively manageAnd remember: AI SDRs scale what works. They don’t fix what’s broken.Get your fundamentals right first. Then let AI take you to the moon.We’ll be covering our RevOps, customer success, and marketing AI deployments in Part 2 next week. You can see all our tools and specific use cases at saastr.ai/agents.Or come see it all in action at SaaStr London in December, where you can interact with our AI agents live and see exactly how we’ve built this. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

Ahead of SaaStr AI London on Dec 1-2 (See you there!) we’re taking a look back at some of our favorite sessions from our European events. It was so great when Olivier Pomel, founder CEO of Datadog, joined us as they crossed $100,000,000 ARR in a candid conversation it would be harder to do today post-IPO.The First $100,000,000 ARR at Datadog: How Olivier Pomel Built a Customer-Centric Monitoring GiantFrom zero lines of code to 700 employees and doubling revenue annually, Datadog CEO Olivier Pomel shares the counterintuitive strategies that built one of the most customer-obsessed companies in B2B SaaSOlivier’s Top 5 Toughest Learnings* You can’t be customer-focused if you’re sales-driven OR engineering-driven - Most companies fall into one trap or the other. Sales teams optimize for closing the next deal (short-term), while engineering teams build for the long-term without bridging back to customers. Customer-centricity requires daily vigilance against both.* Closed alphas with “perfect customers” give terrible signal - Handpicking the best companies and best people for early access actually makes it harder to learn. Customers need to self-select when the timing is right for them. Open betas revealed infinitely more than curated alphas ever did.* Month-to-month contracts are better than annual deals for learning - Every instinct (and investor) tells you to sell annual contracts. But monthly contracts force bad news to surface immediately instead of a year later. A year of going in the wrong direction is devastating for a young company.* There’s no MVP for enterprise infrastructure - The conventional wisdom about shipping minimal products doesn’t apply when selling to enterprises who need comprehensive solutions. You need depth across many features before you’re minimally useful. It’s a continuum, not a single viable moment.* Pricing conversations reveal product truth better than any metric - Putting a dollar amount on features focuses customers’ minds like nothing else. When customers say “I won’t pay for that,” you get brutally honest feedback about value. This friction is healthy and teaches you where to go next.When Olivier Pomel and his co-founder started Datadog in 2010, they didn’t write a single line of code for the first six months. For two engineers itching to build, this took “some restraint,” as Olivier puts it. But this decision to obsessively listen before building became the foundation of a company that would redefine infrastructure monitoring and grow to 700+ employees while doubling in size every single year.At SaaStr Europa, Olivier pulled back the curtain on how Datadog became one of the most customer-centric companies in enterprise software—and why being truly customer-focused requires constantly fighting against your natural instincts.The Problem: When Great Teams Hate Each OtherThe genesis of Datadog came from a painful problem Olivier and his co-founder experienced firsthand. Despite working together across four different companies, knowing each other extremely well, and building their teams from scratch with a “strict no a*****e policy,” they ended up in a familiar nightmare scenario two years in.“We ended up with developers that hated operations, operations that hated developers, fingerpointing—all of the things that you can imagine,” Olivier explained.The question became simple: Why don’t we give all of those teams the same viewpoint? How do we get them aligned and understanding their infrastructure the same way?This became Datadog’s founding mission—bringing DevOps together and bridging the gap between development and operations teams. What they didn’t fully realize at the time was that this wasn’t just a nice-to-have feature. It was actually one of the KEY reasons why companies would migrate from legacy IT to the cloud.“We ended up right in the center of it,” Olivier said. “Today the company is about 700 people. We’ve been doubling the size of the company every single year. It turns out everybody is moving to the cloud and everybody needs to understand what’s happening to their systems and applications.”The Counterintuitive Truth About Customer-CentricityHere’s the part most founders get wrong: you can’t be customer-focused if you let your company become sales-driven OR engineering-driven.“Everybody wants to be customer-focused,” Olivier noted. “But most companies end up being either sales-driven or engineering-driven. If you want to be customer-focused, you can’t be either of those.”The Sales-Driven Trap: Sales teams are phenomenal at figuring out what’s going to get a deal done. But very often, getting the next deal done is NOT what you want to do for the long run for your customers. Short-term thinking dominates.The Engineering-Driven Trap: Let your engineering teams run on their own, and you’ll end up with organizations where people focus way too much on their solutions and way too much on the long term. You’ll struggle to bridge that gap back to the customer.The solution? “It’s a struggle of every day to make sure that we go back to the customer and start from there.”This wasn’t a conscious choice at first—it was forced upon them. Starting in New York (not the Bay Area), without Google or Facebook pedigrees, without millions in funding, and without a suite of VCs telling them they were geniuses, Datadog had no choice but to obsessively focus on the problem.“We basically had to focus on the problem. We spent the first couple of years listening to customers and trying to understand what the problem was.”The First Two Years: Why They Didn’t Write Code for 6 MonthsWhen you’re engineers and you start a company, every instinct tells you to build. Olivier and his co-founder resisted this for six months—and discovered something remarkable.When you don’t have anything to sell, everybody is super happy to talk to you.“You’ll get hours and hours of really fantastic people at fantastic companies and they’ll spend all that time explaining to you what their problems are, what’s working, what’s not working for them. They’ll be extremely candid.”This changes the moment you have something to sell. “Then you’re tainted. You have too much of a vested interest and you’re trying to push something, so people won’t open up so easily.”After six months of research, they spent another six months building their first alpha. When they deployed it to a small number of customers, they noticed the product was “way too open-ended, way too general.”But here’s where they made a critical discovery that contradicts conventional wisdom about customer selection.The Open Beta Revelation: Let Customers Self-SelectDatadog initially ran a closed alpha, handpicking the best companies and the best people at those companies. “It was actually really, really hard to get a lot of signal from these customers,” Olivier admitted.Then they opened it up to a wide beta—and everything changed.“It’s a lot easier for users to self-select and start using your product than for you to understand for whom you’re going to be the right thing at the right time.”You can’t predict when customers will have free bandwidth, when all the stars will align for deployment, or when your solution will be exactly what they need. Let them tell you by trying it.This open approach became core to Datadog’s growth strategy and fed directly into their customer-centric philosophy.Why Datadog Doesn’t Believe in MVPsAsk most SaaS founders about MVPs and you’ll get textbook answers about shipping fast and iterating. Olivier has a different take.“I don’t think there’s an MVP for what we do. I think it’s a myth,” he stated bluntly.Here’s why: Datadog sells to enterprises. They sell a product that needs to monitor everything happening across entire infrastructures and applications. There’s a very large number of features customers rely on, and none of these features in themselves are revolutionary.“You need to have them, otherwise you cannot be minimally useful. It’s more of a continuum—you keep adding those features, and at certain point you have enough of them that customers can start buying.”The hard part? Figuring out which features matter most. “That’s why you have to go back to the customer and make sure you have enough of them so you can actually extract some signal from those conversations.”For enterprise B2B infrastructure companies, the MVP concept often doesn’t apply. You need depth before customers will even consider you.Scaling Customer-Centricity: The 10-100 Person PhaseWhen Datadog hit the market, they were about 10-15 people. The phase of reaching initial scale ran from there to just under 100 people. This is where most companies start to lose their customer focus—but it’s also where you can build the most powerful feedback mechanisms.Strategy #1: Only Sell Month-to-Month Contracts“We didn’t sell yearly deals for a very long time. We only sold month-to-month, meaning customers had the opportunity to churn all the time.”This was structural, not accidental. If something is wrong with the product, if you don’t solve the right problem, if it’s not valuable enough—you know right away. Customers churn and you can have a conversation with them.“If you start selling term deals, you’re going to have the bad news about a year later. By then you’ve wasted a year. You’ve gone in the wrong direction. You’ve made all these mistakes you could avoid otherwise.”The lesson: Find structural ways to get bad news quickly. Don’t let long-term contracts mask product problems.Strategy #2: Put Every Engineer on Support RotationEvery single engineer at Datadog goes on a one-week support rotation throughout the year. Not just professional support engineers—every engineer in the company.“Engineers are super happy when it starts, and they’re even happier when it ends,” Olivier joked. “But really what it gives is empathy for the customer.”Engineers see what problems customers face day-to-day. They see the consequences of their design choices. Sometimes features that seemed like fantastic ideas turn out to be confusing from the customer’s perspective.“It was a great idea, but we’re going to change it because it doesn’t work quite as well.”Strategy #3: Bring Engineers to Conferences to Give DemosDatadog does extensive event marketing because companies migrating to the cloud need to go to conferences to learn. They exhibit at these conferences, give lots of demos—and they bring engineers to do all of it.“Everybody’s on a rotation and everybody’s going to get to one, two, three, five maybe of those conferences and spend every time a full day basically giving demos to potential customers or existing customers and answering questions.”This builds empathy for customers and makes the whole engineering team more confident about the problems they’re solving and their relationship to end users.But here’s the trick: Many engineers aren’t naturally comfortable seeking strangers’ attention at a booth for hours. The solution? Have other people on the team responsible for getting attention and introducing visitors to engineers. Then engineers can settle into a rhythm of demos and questions without the anxiety of having to attract people.What Metrics Actually Matter (and Which Ones Don’t)Here’s something counterintuitive from a company that processes 4-5 trillion customer metrics per day: Datadog isn’t incredibly metrics-driven for their own product.“Most of the metrics are lagging indicators and they don’t convey all the nuance of the value that our customers are going to find with our product. So we don’t actually train ourselves to optimize to the metrics.”What they DO watch:* Volume of data and infrastructure customers are monitoring: A sign of the value they’re providing, because customers deploy them into more places* Engagement: But this isn’t revolutionary* Churn: Which is why they focused on month-to-month contracts early onThe philosophy: Don’t let metrics become a substitute for actual customer conversations and understanding.Scaling to 700: When Everything ChangesAt 700 people across 16 offices worldwide, some of the early strategies don’t work anymore. The executive team can’t meet with all customers all the time. The co-founders and CPO can’t do all the product work.This is where Datadog made a critical hire: product managers.But not just any product managers. They didn’t hire their first PM until they were over 100 people, and when they did, they defined the role very specifically.“Product managers are not here to invent product. They’re here to spend the most time possible with the customers outside of the office.”Their job:* Understand what customers’ problems are* Understand what they’re using, what they’re not using* Understand what they’re paying for and why* Get the best understanding possible of the problem and the value those problems represent* Present the product and get feedback“Their role is not to invent the product but to make sure it solves the right problem and has the right value.”This is a fundamentally different view of product management than most companies have. PMs aren’t visionaries sitting in a room designing the future—they’re customer researchers and feedback conduits.The Startup-Within-a-Startup ProblemWhen you start branching into new products or new markets, you’re suddenly back in startup mode—and it’s easy not to realize it.“It’s easy to get used to winning, and then you expect everything’s just going to work this way,” Olivier explained.Teams working on new products need to behave exactly as you would when you were a tiny company:* Soliciting as much feedback as quickly as possible* Going as wide as possible with betasThere’s tension here. You don’t want to taint the opinion of your existing product by releasing a new product that isn’t as polished. But if you want that new product to work, you need to go very wide very quickly.Most importantly, you need to create a culture where people seek the bad news.“It’s much better to go to a customer and ask them, ‘This new product—are you going to pay for it? You’ve tried it. Is it good enough? Are you going to buy it?’ It’s actually much better to hear from them, ‘No, I’m not going to buy it, and this is why,’ than to kick the can down the road and hope that maybe in 3 months, 6 months, 9 months they’ll change their mind.”When you know what’s not working, it gives you a thread to pull. It tells you where you need to go.The Counterintuitive Truth About PricingHere’s where customer-centricity gets tested: the pricing conversation. Customers want to spend as little as possible. You want them to pay as much as possible. How do you stay aligned?First, agree on what kind of company you are. Olivier sees two types:Type 1: Low-End Disruptors The major feature is that it’s cheaper. The dynamic with customers is they’ll push you to be cheaper and cheaper and to figure out what you can take out of your product so it can be as cheap as possible.Type 2: High-Value Products You charge more for your product and you have to deliver more value for that. The cycle with your customer is to deliver more and more.“You have to agree with your customer that they’re looking for high value, high impact—not as cheap as possible—because otherwise you’re never going to meet.”Once you’ve agreed on that, you’re aligned in the mid and long term. Your customers want you to be successful. They want your company to be in business two years from now. They want you to ship new features. They really want you to be successful.After that, there’s still friction—but that friction is how you learn about the value of your product and where you need to go next.“Asking for money and putting a dollar amount on the product really focuses the mind for the customers. It helps you get really, really good feedback on where you need to go. In the long run, it’s a very healthy relationship to have.”The key insight: See your customer as a partner. Frame it as “Let’s do this together. You want me to succeed and I want you to succeed, and the commitment on your side is to pay.”How Olivier Maintains Culture at 700 PeopleCulture isn’t about writing down seven values on a wall (though Olivier says they’ll do that eventually as they continue to scale). Culture is really about who you hire, who you fire, and who you promote.To maintain a customer-centric culture at scale, Datadog trains managers to care about the details. Olivier himself still sees every single customer complaint and support email that comes in.“I don’t read most of them. I delete most of them right away. I just scan them very quickly. But what this does is it helps me pattern match. It gives me a sense of what’s actually happening, how people are actually reacting to the product, what they’re saying.”He never acts on them directly. He never goes back to the customer or tries to solve the problem. But if he wants to affect change, he goes back through the management chain to make sure people get feedback and decide whether they need to change something.The other critical element: Be super careful about how you talk about customers inside the company.In sales, it can be tempting to talk about customers in manipulative ways as you try to manage them through a process. In engineering or product, you can think customers are making the wrong choices—”Why are they doing this stupid thing with our product?”“The way we see it is if they’re doing a stupid thing with our product, it’s our fault. We let them do that. We made them think they should be doing that, or we didn’t explain it right.”Always assume the customer is right. And even if you think they’re wrong, the fact that they’re thinking something wrong is itself a fact. It exists. You have to deal with it and you have to value it.You can’t let yourself dismiss what you hear from customers.The Acquisition Test: Optimizing for After the DealDatadog has made two acquisitions, both heavily tilted toward the tech platform, product, and team—not revenue streams to add to their sales engine.“What we optimize for is what’s going to happen after we close the deal. Are people going to stick around? How long are they going to stick around? Are we going to be able to build on top of that team and turn that team of 20 we just acquired into a department of the company that’s going to have 200, 500 people?”It’s all optimized on the fit and what happens after the acquisition—whether the companies share the same values.The One Piece of Advice: Run Toward the Bad NewsIf there’s one principle that runs through everything Datadog does, it’s this:“Really look for the bad news. Run toward the bad news.”Whether you just shipped the product or you’re renewing a big customer, the wrong thing to do is show up in front of the customer hoping they won’t bring up the issue in the account. “Hey, maybe with a bit of luck we can get away with it.”“Actually no—the first thing you do is you talk about it. You bring it up. Then you hear directly from the customer what they have to say about it. It actually goes a long way toward building the partnership, and it also helps you learn from the conversation with your customers.”This applies at every stage:* First six months: Don’t build—just listen to bad news about the market* Alpha stage: Open it up to get bad news faster* Launch phase: Only sell month-to-month so bad news comes immediately* Scale phase: Put engineers in support so they hear bad news directly* New products: Ask customers point-blank if they’ll pay for it* Big renewals: Lead with the problemsThe Bottom LineBuilding a customer-centric company isn’t about being nice or having good intentions. It’s about building structural mechanisms that force bad news to surface quickly and loudly.It’s about resisting the natural pull toward being sales-driven (short-term thinking) or engineering-driven (disconnected from reality).It’s about spending six months not coding when you’re itching to build.It’s about letting customers self-select rather than trying to control who uses your product.It’s about putting engineers in uncomfortable situations where they have to face customers directly.It’s about only selling month-to-month when your investors want annual deals.It’s about asking customers if they’ll pay for something and being ready to hear “no.”From zero to 700 people and doubling revenue every year, Datadog’s growth came from one counterintuitive principle: The best way to win is to constantly seek out why you might be losing.“You have to the wrong thing to do is hope the customer won’t bring up the issue,” Olivier concluded. “The first thing you do is you talk about it. You bring it up. Then you hear directly from the customer what they have to say about it.”That’s customer-centricity at scale.Olivier’s Top 5 Mistakes (In His Own Words)* Running a closed alpha with handpicked customers - “It was actually really, really hard to get a lot of signal from these customers.” Trying to control who got early access and picking the “perfect” companies backfired. Opening up the beta was when everything changed.* Building the first alpha too open-ended and general - “We noticed that the product was not exactly what it needed to be. It was way too open-ended, way too general.” Even after six months of customer research, they still got the scope wrong initially.* Not realizing how big the cloud opportunity would be - “We didn’t quite understand at the time that the cloud was going to be so big. We thought, hey, looks cool, why don’t we build it for companies moving to these new cloud environments.” They underestimated the magnitude of the shift they were riding.* Not understanding that DevOps bridging was the killer feature, not just a feature - “We didn’t quite realize was that bringing DevOps together, bridging those two teams, was not just a feature of the new world. It was actually one of the key reasons why companies would be moving from legacy IT to the cloud.” They had the solution before they understood its full strategic value.* Almost letting new products follow the old playbook instead of startup mode - “It’s easy to get used to winning and then you expect everything’s just going to work this way.” When launching new products at scale, they had to consciously fight against the temptation to skip the scrappy early-stage customer validation work that made them successful in the first place.Want more insights on building customer-obsessed B2B + AI companies? Join us at the next SaaStr AI London Dec 1-2, where founders and CEOs share the unfiltered truth about scaling from $0 to $100M ARR and beyond.Thanks for reading SaaStr AI: How To Sell, Scale, and Win! Subscribe for free to receive new posts and support my work. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

We’re back! Harry, Rory and Jason!The venture capital playbook is broken. Not bent — broken. In the latest 20VC x SaaStr episode, Harry Stebbings, Jason Lemkin, and Rory O’Driscoll dissect why even Sequoia is making dramatic leadership changes, why seed investing at $50M pre-money might not work anymore, and what it actually takes to build venture returns in the age of AI.This isn’t your typical venture conversation about “exciting trends.” This is three investors with $3+ billion in combined AUM telling you what’s actually working, what’s spectacularly failing, and why the old playbook from 2015-2022 is now a liability.Key TakeawaysOn Venture Capital Evolution:* Sequoia’s leadership transition reflects broader industry truth: most VCs and executives from the last decade aren’t the right people for the next decade* The pace of AI evolution means knowledge from 6 months ago is probably wrong; staying current requires dedicated time investment* Partnerships are inherently dysfunctional when performance can’t tie to economics, creating inevitable internal tensionOn AI Investment Strategy:* Only three ways to win: (1) Attach to compute budgets, (2) Replace human headcount, or (3) Massively displace incumbents* Using AI to “make your product better” no longer earns any kudos — that’s table stakes in 2024* Co-pilots were the 2024 story that didn’t work; agents becoming actual team members is the 2026 opportunityOn Deal Dynamics:* Getting into deals at 5-10M ARR requires top-decile metrics — there’s almost no middle class of fundable companies* The quality and speed of competitive clones has increased dramatically, compressing the window for building moats* Traditional seed defensibility is dead; founders must run faster and bet on scale creating the moat, not early product advantagesOn Portfolio Construction:* With increased variance in AI deals, diversification becomes more critical, not less* Small fund sizes ($40-100M) with acceptance of dilution can generate superior returns (10x+) versus large funds maintaining ownership (5x)* 80+ company meetings per week per partnership is one approach; building deep relationships with fewer founders is anotherOn Fundraising Process:* The best fundraises don’t feel like processes — they’re cultivated over months with 3-4 investors ready before the data room opens* Taking a term sheet immediately versus “running a process” depends on capital efficiency and relationship quality* Founders often overlearn “run a process” advice without understanding the optimal approach is having everyone ready to commit before you formally raiseOn Market Dynamics:* Companies attached to AI compute infrastructure (like DataDog) are crushing it; those just using AI for product improvement (like Duolingo) are getting punished* The $100M ARR milestone with 50 people (like Gamma) represents a new efficiency paradigm* Capital-efficient outcomes (Billion to One at $5B, Navan at $4.5B) deliver superior investor returns despite smaller headline valuationsSequoia’s Move: What It Really MeansWhen Sequoia replaced Roelof Botha as managing partner after just three years with Pat Grady and Alfred Lin in a split leadership structure, the venture world noticed. But the reaction from our panel was telling: this wasn’t about interpersonal drama or normal succession planning.“Whenever you have a CEO change happening in venture, it’s because something isn’t working,” Rory stated bluntly. “This is dissatisfaction about how the firm is doing relative to the competition. They missed some rounds and some deals. They passed on some great companies.”But rather than viewing this as a Sequoia-specific challenge, the panel sees it as symptomatic of a broader industry issue.“More people should be stepping aside today,” Jason argues. “VCs, executives, founders. Most folks from the last decade or 15 years are not the right people for the next decade. I genuinely believe it across my ecosystem. I struggle to even recommend a lot of the CROs and executives I know for roles today.”Rory adds context: “The pace of evolution is so fast. If you decide what I knew 6 months ago is still useful, you’re probably going to be wrong very quickly. That’s what I find the most stressful about right now.”The meta-lesson? Even the best firms are struggling to keep pace with AI, which means everyone else should be asking themselves hard questions about whether they’re still the right people for this moment.And Sequoia gets credit for one thing: ruthless decisiveness. “They did not do that fatal error of saying it’s someone’s turn so we’ll leave him in,” Rory notes. “They ruthlessly said, if we’re going to compete, we need these people, not those people.”Michael Burry’s $1.1B Bet Against Nvidia: Why Shorting AI is Brutally HardMichael Burry made headlines (again) by shorting Nvidia and Palantir to the tune of $1.1 billion. Unlike most pundits who just discussed whether he’d be right, Jason actually did the math on what it takes to make money on these bets.“It brings home how hard a business it is to bet against AI capex,” Jason concludes. “You not just got to be right, but you got to be right on timing.”Rory’s take: “It’s easy to be roughly right. It’s very hard to imagine that the AI capex boom doesn’t have a significant correction. But going from that kind of armchair podcast statement to actually being able to make money on it — that’s damn hard.”The counterpoint to all this short-selling cynicism? The revenue is actually showing up. OpenAI is projected to hit $20B ARR this year, while Anthropic projects $70B ARR by 2028. The growth rates aren’t just holding — they’re accelerating.Harry drives this home: “Are we being overly British? Are we looking for a problem that’s not there? The revenue is showing up in billions.”The lesson: It’s intellectually easy to be a skeptic about AI valuations. It’s financially hard to make money being that skeptic. And meanwhile, the actual revenue keeps proving the bulls right.Gamma’s $100M Revenue Run Rate with 50 People: The New Efficiency ParadigmFew stories capture the AI revenue opportunity better than Gamma’s latest round. The company raised $100M at a $2.1B valuation, having hit $100M in revenue with just 50 people and 2 million users.Jason and Team SaaStr are super-users and he walks through exactly how they use it at SaaStr: “Instead of sending sponsors the same dated prospectus, Gamma automatically pulls all of our data from Salesforce and our marketing automation system. It knows the exact number of leads and ROI. It knows who their competitors are and makes a fully dynamic piece of collateral in about 10 minutes.”Here’s the math that’s important: SaaStr pays $100/month for Gamma — $1,200 per year. That’s stealth TAM expansion. “How much do we spend for Google Slides? Zero. It’s built in. How much do we spend on PowerPoint? I don’t even know where my key is to Microsoft Office. So it is a stealth TAM expansion.”But Jason’s most important insight is about what’s coming next: “The lame thing about co-pilots is they were just tools. When the AI is part of your team for real, not VC talk, the amount of revenue that it’s accessible is so high.”He distinguishes between AI as a tool versus AI as a team member: “It is sufficiently autonomous, knowledgeable and powerful to complete material high-value tasks on its own with some daily discussions just like on our team. The level of autonomy and capability — Gamma, go into my Google Calendar, create prospectuses and sales collateral for all 20 sales calls this week, pull all the data from Salesforce and HubSpot and Marketo, review once, and distribute to the team.”The implication: We’re transitioning from AI making us more efficient (2024) to AI being actual members of the team (2026). That’s where the revenue explosion happens.But there’s a warning embedded in Gamma’s story too. Canva now has a Gamma clone that’s “pretty good.” The competitive response time has collapsed from years to months or even weeks. This leads to one of the episode’s most important debates…Does Defensibility Exist Anymore? The Most Important Question in VentureThis might be the single most important strategic question facing every seed and Series A investor right now. And the panel doesn’t fully agree on the answer.Jason frames the problem: “The quality of clones is only going up. I can think of one investment I’ve made that has had five clones in the first 30 days, including one from a cloud leader. The ability of AI to enable us to clone better stuff faster — Canva is borderline competitive with Gamma and wasn’t when Cliff was on the show. What does seed investing mean when you might see 10 better versions in 30 days?”Rory’s response: “I don’t think you can have a major defensibility at the seed or even frankly the stage we’re investing at. The defensibility theorem emerges at scale. Once you become the anointed winner, once a market coalesces and there’s two or three people, you know, it’s yours to lose.”But Rory pushes back on the idea that this makes seed investing impossible: “You just have to internalize the game you’re in. Awesome team, run fast, be superlative on technology, get your distribution early, and then rely on that. You can’t be anointed the winner up front. Get over it everybody.”Jason’s concern is economic: “Is that okay at 50 post for a seed round? Do the outcomes justify it? If I’ve got to spread $5 million checks around at 50 post, it’s tougher.”Harry asks the critical question: “Are we getting paid for the risk?” He notes that even at the Series B stage, it’s unclear who will win in many categories. Looking at customer support AI companies that have raised Series B rounds: “I got no idea who the winner is in that category and I don’t think anyone does to be honest.”The debate crystallizes around whether you can know enough at the B to justify the valuation:Rory’s view: By the Series B, you can see rank order. “Once the horses are running and once they round the first furlong, you can actually see the rank order of where they’re running in a way you can’t at the early A.”Harry’s counter: Looking at codegen — even with clear leaders like Cursor emerging, you have Codex making incredible ground, Claude Code advancing, and Replit and Lovable coming from different angles. “I think that’s still entirely up for grabs.”The practical implication: This uncertainty means one of two things:* Accept that seed/Series A investing requires higher diversification (40+ companies instead of 20)* Or accept lower ownership percentages betting on outcome expansionJason’s conclusion: “Maybe you need 40 deals at $5M. That’s $200M for first checks. $200M for reserves. That’s $400M. $100M for fees and backup. I need at least $500M for my little seed fund to make the math work.”The Three Ways to Actually Capture AI RevenueAmid all the complexity, the panel crystalizes a simple framework for what actually works in AI investing right now.Path 1: Attach to Compute InfrastructureDataDog crushed earnings, with stock up 23% and $15M+ in AI-native customers. Why? They’re selling to the people making AI, and as those companies grow, DataDog sells more too.Jason: “The AI leaders, the hyperscalers — they’re starting to buy like classic B2B companies. They’re recycling the same people in procurement. So if you’re attached to AI budget and you’re a DataDog era, you’re actually going to have a great 2026.”Rory adds: “DataDog is a core piece of compute infrastructure and these hyperscalers are the most compute-intensive companies that have ever been known. If you’re selling compute stuff, you should be having a great quarter. If you’re selling routers, switches, interconnects, whatever it takes to stand up Stargate — you’re going to be golden.”Path 2: Replace Human Headcount“Where are you replacing humans for real?” Jason asks. “Where are you going to go in and reduce the headcount that vendor needs by half?”This is the Replit story — Jason deployed 20+ AI agents that generate significant revenue while requiring substantial daily management but far less human headcount than traditional approaches.The key distinction: “Most of the time, ‘replacing humans’ is a story. But when it’s real, when the AI agent is sufficiently autonomous to complete material high-value tasks, that’s when you’re actually part of the team, not just a tool.”Path 3: Massively Displace an Incumbent“You have a third option,” Jason concedes. “Use AI to massively displace an incumbent and steal all the revenue. In fact, that’s the history of B2B software mostly. I just don’t know how many of our public leaders are in a place to steal their own revenue.”But he’s skeptical this is as compelling as the first two: “There’s like 400 AI CRM startups out there all saying they’re going to eat HubSpot’s and Salesforce’s lunch. That’s not exciting to me as an investment. That’s much less exciting than truly replacing 90% of your GTM team.”If you’re not doing one of these three things, you’re probably not going to see explosive growth. Using AI to make your product better? That’s table stakes now, not a competitive advantage.How to Actually Run a Fundraising Process (Without Running a Process)One of the most tactically useful segments came when Harry shared a founder interaction that frustrated him: the company said they wanted to “run a process” despite Harry offering terms they’d stated they wanted.Harry’s reaction: “I said, listen, if you’re running a process, you’re either optimizing for price or partner selection. I’m giving you a great price today that you said you wanted, which means you’re not optimizing for price. Which is just saying you think you can get better than me. In which case, just tell me straight.”Rory sees both sides: “From the founder’s side, they’re correct. I see more failed financings because they didn’t run a process than because they did.”But then he makes a critical distinction: “What typically happens when people ‘don’t run a process’ is someone comes in, says they’re interested, and the founder shares information serially to someone not yet ready to commit. That’s a mistake. That’s running an accidental process.”Jason has the most nuanced take on optimal strategy: “The best founders cultivate enough interest with enough good VCs that if they hit the number, they just send an email. Harry, I’m thinking about raising around before the end of the year. And Harry could say, I’ll give you a term sheet today. And the right answer is: ‘I love you, Harry. I’m not ready today. I will be ready at the end of the year.’”The key insight: “The optimal way to run it is for everyone already to want to invest for real without games before you open your data room. The best run processes don’t require a data room. Not a traditional one. Only one for diligence.”Rory synthesizes: “The best run processes don’t feel like a process, but they are. If a founder is smartly nurturing relationships, keeping people broadly informed, then tries to time the interest such that when they’re ready to put their hand up, there’s three people who are primed and ready to go — that is perfection itself.”But here’s the brutal caveat: Jason points out that this only works if you have stellar numbers. “You’re either YC, Neo, South Park Commons, got something, get funded — or who the hell is going to find you? You better be hot AI-native with top-quartile venture growth or you ain’t getting funded.”Harry shares a data point that proves this: A company that grew from $400K to $3M ARR (classic enterprise SaaS, 10x growth) had 120 meetings and got one term sheet at $10M on $40M post. That’s 12x revenue for a 10x grower — which five years ago would have generated five term sheets from top firms.“It’s the most binary fundraising environment in our lifetimes,” Jason concludes. “You’re either Captain Obvious getting funded, or you’re not seeing it anywhere.”Duolingo Down 25%: The Cautionary Tale of the “Wrong” Kind of AIWhile DataDog surged 23%, Duolingo crashed 25% in the same week. The contrast illustrates everything about what’s working and what’s not in AI.Jason’s diagnosis: “Duolingo has the ‘wrong’ kind of AI. Duolingo is using AI to make its product better. Hooray. Every single portfolio company at scale should be using AI by this point to make your product better. This is not 2023. You don’t get any kudos for sprinkling AI dust on your product.”Rory pushes back slightly on behalf of consumer apps: “If you’re an infrastructure company, you can co-attach to compute. If you’re a new AI apps company, you are using that compute. But if you’re like Duolingo — preexisting — what you can do is co-adopt the technology.”But Jason sees a bigger problem: “Duolingo took money from Berlitz and all these language schools. Hooray, you did that. Now, where are you going to disrupt humans? This is your job. You already disrupted those humans. Unfortunately, they’re gone. Where’s the next level of human disruption?”Rory makes an interesting point about where the AI opportunity in education actually is: LLM learning is equivalent to one-on-one human tutoring. What AI should enable in education is allowing everyone to get one-on-one learning instead of group-based learning. But there’s no budget in K-12 to give every kid a customized tutor. The market is adults wanting to learn a second language for business who would pay for a one-on-one coach — a niche market.Jason’s framework applies here too: “If you’re not removing humans from the equation, you’re going to be heavily discounted. You’re either getting money from compute or you’re getting money because you’re using AI to replace humans. Otherwise, congratulations on your 14% growth.”Portfolio Construction: The Math Actually MattersOne exchange revealed how different the panel’s approaches are to portfolio construction — and why both can work.Harry revealed his team’s meeting load: “With my four investing partners, we have 80 net new companies that we meet in person per week.”Jason’s response: “I would resign. I would give you all my carry back. I sold my companies because I didn’t want to spend my life in meetings. I’ll do one a week, two max.”But Jason’s critique cuts deeper than meeting volume: “When you do 200 deals between your team, you got to meet 500, 600, 700 founders a year. I don’t want to meet 500 founders a year. Unless you’re great, you’re going to blow my mind. But if you’re not, I’m gonna start yawning about 15 minutes into this meeting.”Rory defends the high-volume approach: “Even on an okay deal, you learn nuances about a specific market from the one-on-one that you wouldn’t get from the presentation. The person inside living it every day has a crucial kernel of knowledge that you just can’t access any other way.”But the conversation reveals a deeper truth about fund construction:The Small Fund, High Multiple Approach (Hummingbird-style):* Keep fund size at $40-100M* Accept dilution in follow-on rounds* Start with 20% ownership, end with 12%* Generate 8-10x returnsThe Large Fund, Maintained Ownership Approach (Lightspeed-style):* Raise $250M+ funds* Deploy significant follow-on capital to maintain ownership* Generate 5x returns but on much larger capital baseRory’s key insight: “Both are great outcomes. They’re just different ways to play the game. The interesting thing is both outcomes have great outcomes for the GP. But if you only have $1 to play with as an LP, you want the small fund 10x. If you have to deploy $100M, you have to do the big fund. The high small fund accept the follow-on dilution but make a marvelous return is the compelling product for the marginal dollar.”The Hummingbird Outlier: How to Actually Generate 10x ReturnsSpeaking of Hummingbird, their Billion to One investment deserves its own section because it represents what’s possible when everything goes right.Hummingbird did their first bio deal in Billion to One and now has an $800M position on the IPO — likely representing an 800x+ return on their initial investment from a sub-$100M fund.Harry marvels: “You want freaking great venture returns? Credit due. Amazing.”But Jason asks the tactical question every seed investor should be asking: “How did they collect the capital as a seed manager to deploy enough to maintain the ownership? I don’t see how I’ll ever own 18% of something at IPO ever again.”The answers are telling:* Capital efficiency — Billion to One didn’t require massive rounds* Follow-on investment — They did put in subsequent checks* Concentration — They bet big on the winners* Accepting some dilution — Rory suggests they may have gone from 20% to 12% but on a massive outcomeRory makes the crucial point: “You can accept some dilution and optimize for multiple rather than ownership. If you put in $4M and get diluted from 20% to 12% but it’s a $5B exit, you’re a hero.”Jason counters with what’s exciting about capital efficient outcomes: “Navan at $4.5B, Billion to One at $5B — the power of capital efficiency and running lean delivers superior investor returns despite smaller headline valuations.”The lesson: Small funds generating 10x+ returns by accepting dilution but betting on capital-efficient companies may be the optimal strategy for seed investors who don’t have the deployment capacity of mega-funds.What This All Means: The Real State of Venture in 2025Pulling all these threads together reveals several uncomfortable truths:1. The Middle Class of Fundable Companies is Gone“It’s the most binary fundraising environment in our lifetimes,” Jason states flatly. The classic SaaS triple-triple-double-double company? A company growing from $400K to $3M ARR would have had five term sheets from good firms. Now: 120 meetings, one term sheet.2. Most People from the Last Decade Should Step AsideThis isn’t ageism — it’s pace of change. “The most folks from the last decade or 15 years are not the right people for the next decade,” Jason argues. “I struggle to even recommend a lot of the CROs and executives I know for roles today. We don’t even need half the VCs we have today for the AI world.”3. Diversification Matters More, Not LessWith increased variance in outcomes and compressed time to clone, the math says diversification is critical. But diversification at seed prices requires massive funds or acceptance of lower ownership.4. There Are Only Three Ways to WinAttach to compute budgets, replace human headcount, or massively displace incumbents. If you’re not doing one of those three things, you’re probably building a slow-growth SaaS company in a world that doesn’t want those anymore.5. Capital Efficiency Delivers Superior ReturnsBillion to One at $5B and Navan at $4.5B generated better investor returns than most $20B+ outcomes because they raised less capital. The lesson: optimize for investor returns, not headline valuations.6. The Best Fundraises Don’t Feel Like ProcessesCultivate relationships over months. Have 3-4 investors ready to commit before you formally raise. Send one email and get multiple term sheets. That’s the optimal path.7. If You’re Not Excited, You Should RetireJason’s most important point: “If this isn’t the most exciting time of your lifetime, you’re doing it wrong. This is the first time software has gotten better since the three of us met. If you’re not incredibly excited, retire. No shame in that. Put the rest into NASDAQ — you’re going to make more than most VC funds anyway.”Looking Forward: What 2026 BringsThe panel is clear about what’s coming: “This year was AI works. Next year is AI is part of your team.”That transition — from AI as a tool making you more efficient to AI as an actual member of your team with autonomy, knowledge, and capability — is where the revenue explosion happens.Jason’s final thought captures the optimism underlying all the tough talk: “If you’re not incredibly excited right now, if this isn’t one of the most exciting and stressful times simultaneously, you’re missing it. This is a great enormous mega trend. It’s the biggest mega trend maybe since the early days of the internet. Leaning into it is the only sensible thing to do, and playing against it is dumb as rocks.”The old playbook is broken. The new playbook is being written in real-time. And the investors who figure it out — whether through small concentrated funds like Hummingbird or large diversified portfolios like Harry’s approach — will capture the extraordinary returns available in this moment.Everyone else? Maybe it’s time to buy that beach house.Quotable MomentsOn Venture Evolution:“The most folks from the last decade or 15 years are not the right people for the next decade. I genuinely believe it across my ecosystem. I struggle to even recommend a lot of the CROs and executives I know for roles today.” — Jason Lemkin“The pace of evolution is so fast. If you decide what I knew 6 months ago is still useful, you’re probably going to be wrong very quickly. That’s what I find the most stressful about right now.” — Rory O’DriscollOn AI Investment Strategy:“Tools are great when the AI is part of your team for real, not VC talk. The amount of revenue that it’s accessible is so high.” — Jason Lemkin“Sell to the people who are making AI and if they grow, you’ll sell more too.” — Jason LemkinOn Defensibility:“You just can’t take that early first month explosion as seriously as you used to. It’s not as defensible.” — Harry Stebbings“I don’t think you can have a major defensibility at the seed or even frankly the stage we’re investing at. The defensibility theorem emerges at scale.” — Rory O’DriscollOn Market Timing:“It’s easy to be roughly right. It’s very hard to imagine that the AI capex boom doesn’t have a significant correction. But going from that kind of armchair podcast statement to actually being able to make money on it — that’s damn hard.” — Rory O’DriscollOn Fundraising:“The best run processes don’t require a data room. Not a traditional one. Only one for diligence.” — Jason Lemkin“The best run processes don’t feel like a process, but they are.” — Rory O’DriscollOn Current Reality:“It’s the most binary fundraising environment in our lifetimes. You’re either getting funded or you ain’t seeing it anywhere.” — Jason LemkinOn Taking Action:“If this isn’t the most exciting time of your lifetime, you’re doing it wrong. If you’re not incredibly excited, retire. No shame in that. You had a great run. Put the rest into NASDAQ — you’re going to make more than most VC funds anyway.” — Jason LemkinListen to the full episode of 20VC x SaaStr on YouTube or your favorite podcast platform. For more insights on venture capital, AI investing, and B2B metrics, visit SaaStr.com and subscribe to the SaaStr newsletter.Thanks for reading SaaStr AI: How To Sell, Scale, and Win! Subscribe for free to receive new posts and support my work. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

By the end of Q3, we’ll had 10 distinct AI agents running in production at SaaStr. 20 if you including less critical ones. Not as a tech experiment or marketing stunt, but as core members of our revenue and operations team.The lineup looks like this:Revenue Team:* 3 AI SDRs handling each of ticket inquiries, sponsor outreach, and sales support (these are different workflows, training, etc)* 2 AI BDRs qualifying inbound leads and nurturing prospects through our funnel* 1 AI RevOps agent tracking and managing our partner pipelineOperations & Experience:* 1 AI Support agent handling event logistics and attendee questions* 1 AI Content Review agent vetting speakers and session proposals* 1 AI Matchmaking agent connecting CEOs and executives at our eventsCommunity & Education:* 1 AI Mentor (SaaStr.ai) providing 24/7 guidance to our community. Try it, it’s free!And we’re not done. The pipeline has 3-4 more AI agents in development.The Operational Reality: It’s A LOT More Work Than You ThinkHere’s what nobody tells you about AI agents in production: they require daily management and review. Not weekly check-ins. Not “set it and forget it” automation. Daily.Every morning, I’m reviewing:* Conversation quality scores from our AI SDRs* Lead qualification accuracy from our BDRs* Edge cases that required human escalation* Performance metrics across all agents* Training data updates and model refinementsEach agent needs constant fine-tuning. The AI SDR that handles sponsor inquiries needed 47 iterations to stop being too aggressive on pricing discussions. Our AI Support agent had to be retrained three times to properly escalate VIP attendee issues.The truth? Managing 10 AI agents is like managing a team of 10 very capable but very literal junior employees who need explicit instructions for everything.But Here’s Why We’re All-In: The Advantages Are UndeniableDespite the management overhead, these AI agents deliver something human employees simply can’t:They never quit. Zero turnover. No recruiting cycles. No onboarding new SDRs every 18 months because they got poached by a competitor offering $10K more.They work weekends. While your human BDR is at Coachella, our AI BDR is qualifying leads and booking demos. Saturday morning inquiries get responded to in under 2 minutes, not Monday afternoon.They don’t complain. No “this lead quality sucks” from the AI SDR. No “I need more training on the new product features” requests. They just execute.They aren’t distracted. Our human SDRs were spending 30% of their time on side hustles, online courses, or job searching. The AI agents? 100% focused on converting prospects and supporting customers.They scale instantly. Need to handle 3x more sponsor inquiries during SaaStr Annual planning? The AI RevOps agent doesn’t need additional headcount approval or three weeks of hiring. It just scales.The Product Knowledge Advantage: They Know Everything “Cold”This might be the biggest unexpected benefit: AI agents know our products, processes, and pricing cold.Human SDRs need 3-6 months to really understand our event portfolio, sponsorship packages, and community offerings. Even then, they’re guessing on edge cases or scrambling to find answers. In fact, most of the SDRs we’ve had never really understood our products at all.Our AI agents? They have perfect recall of:* Every sponsorship package and pricing tier* Historical attendee data and ROI metrics* Speaker requirements and content guidelines* Event logistics for 12+ annual events* Community membership benefits and upgrade pathsWhen a prospect asks our AI BDR “What’s the difference between your Growth and Enterprise sponsorship packages for companies doing $50M ARR?”, it delivers a perfect answer in 30 seconds. No “let me check with my manager” or “I’ll get back to you”. Or worse, no making things up. (Our AIs are well enough trained that hallucinations are minor at best now.)The Financial Reality: ROI Happens Faster Than ExpectedThe numbers are becoming undeniable:Cost per agent: ~$200-4,000/month (including platform, training, and management overhead) Cost per human equivalent: ~$8,000-$12,000/month (salary, benefits, management, office space)But the real ROI drivers:* Response time: Average first response dropped from 4.2 hours to 1 minute* Lead qualification: 67% more leads properly scored and routed* Weekend coverage: 23% of our best leads come in outside business hours* Consistency: Zero “bad days” or emotional decision-making affecting prospect experienceOur AI SDR team has generated $340K in sponsor pipeline so far in Q3 alone, and the quarter has just begun. At a fully-loaded cost of ~$10K/month for all core agents.What Folks Get Wrong“Mistake” #1: AI agents can’t replace human creativity and relationship-building. For complex enterprise deals and strategic partnerships, humans still close. But also, way too many in sales overestimate their skills here. Being a “people person” is not enough.“Mistake” #2: Underestimating the management overhead. Many of you will need to hire a dedicated “AI Operations Manager” role to keep everything running smoothly. RevOps and MarketingOps are becoming radically different positions, requiring different skill sets, in the Age of AI. If you’ve had a bad experience with an AI SDR tool, it’s probably because you expected to buy-and-ignore it. Doesn’t work that way.“Mistake” #3: Some prospects still prefer human interaction for high-value conversations. Although fewer than expected. Be transparent about AI involvement upfront. Many folks are happy to receive a >greatvalueThe Bottom Line for B2B LeadersAI agents aren’t replacing your entire revenue team. But they’re becoming essential for:* Top-of-funnel lead management* 24/7 customer support and qualification* Operational tasks that require perfect consistency* Scaling during peak demand periodsThe companies that figure this out in 2025 will have a massive operational advantage by 2026. The ones that wait for “better technology” or “clearer ROI” will be playing catch-up with teams that never sleep and never quit.Our prediction: By SaaStr Annual 2026, the highest-performing SaaS companies will have AI agents handling 40-60% of initial prospect interactions. The question isn’t if this happens, but how quickly you can operationalize it without breaking your customer experience.Start with one agent. Master the management overhead and the training first. Then scale.Thanks for reading SaaStr AI: How To Sell, Scale, and Win! Subscribe for free to receive new posts and support my work. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

Medha Agarwal is a Partner at Defy VC, where she focuses on investments in AI-first and vertical SaaS companies. She shares insights at SaaStr AI Summit 2025 from the front lines of AI-first product pricing, exploring why traditional SaaS models are declining in favor of transactional pricing, how to choose the right pricing structure for your business, and strategies for capturing value from labor budgets instead of software budgets.Top 5 Takeaways* Transactional pricing is replacing traditional SaaS at an accelerating rate. The fundamental shift is driven by AI’s ability to complete tasks end-to-end, enabling companies to sell into labor budgets rather than software budgets. This opens up significantly larger TAMs that were previously dominated by human labor costs.* There’s a 1.5-2.5x revenue multiple premium for SaaS models in public markets, but transactional models capture more value. While SaaS offers predictability and better cash conversion cycles with annual upfront payments, transactional pricing allows you to scale revenue with customer growth without being constrained by seat count. The trade-off is revenue predictability versus value capture.* Hybrid models are emerging as the best of both worlds. Companies are mitigating transactional pricing unpredictability by implementing tiered subscriptions with usage minimums and overage billing. This provides baseline revenue predictability while maintaining the ability to capture value at scale through consumption-based pricing.* Your pricing model choice depends on four critical factors. Frequency of usage, magnitude of cost savings, workflow integration point, and customer budget type all determine whether fixed-cost, input-based transactional, output-based transactional, or hybrid pricing makes sense. High-frequency tools like Slack need flat fees because users can’t mentally track per-use costs.* Never compete on price alone when entering a market. Undercutting competitors on price creates a dangerous dynamic where you attract price-sensitive customers rather than best-fit customers, leading to higher churn and false signals of product-market fit. Price at par with competitors and win on value, or you’ll be forced into a race to the bottom.The Fundamental Shift: Why SaaS Pricing Is DyingWe’re witnessing a massive sea change in how AI-first products are priced. At DEFY, we’ve seen a dramatic increase in the frequency of transactional-based pricing models. Traditional SaaS pricing, while still popular, is declining rapidly in favor of transaction-based approaches.The reason is straightforward. With AI, there’s an increasing ability for software to complete tasks end-to-end. Some AI businesses are enhancing humans and making them significantly more productive than ever before. Others are eliminating the need for humans to perform many entry-level tasks altogether.This creates a game-changing opportunity. It’s much easier today than it has been in the past to sell into labor budgets, which represent a much larger line item for most companies’ cost structures. Labor is also seen as more mission-critical than software spend. These transactional models can target much larger total addressable markets that were previously captured almost entirely by human labor expenditures, which are now starting to be displaced by software in certain categories.The Three Pricing Model ArchetypesWhen we looked at the pricing landscape for AI-first products, we identified three high-level model types that most companies fall into.Fixed Cost Models: Traditional SaaSThis is your typical traditional SaaS-based, seat-based pricing model. It can be seat-based or location-based pricing. There are compelling reasons why this model has dominated for so long.If you look at the public markets, SaaS models command a 1.5 to 2.5 times revenue premium multiple compared to transactional-based models. There are several reasons for this valuation gap.First, it’s extremely predictable for the company selling the software. Second, it’s very predictable for the customer buying it. They can plan for that spend on an annual basis with high confidence. Third, the cash conversion cycle is excellent. These contracts are often paid yearly upfront, which helps fast-growing companies manage their cash flow effectively.But there are significant trade-offs with fixed pricing models.As companies grow rapidly, their usage of a product often scales much more quickly than the number of seats they need. This happens because they’re not growing headcount proportionally since the tool is making them much more efficient. The result is that the value a vendor is able to capture is much less than the actual value they’re providing to the company.Second, there are use cases where you could add enormous value, but the usage isn’t consistent on an everyday basis. This makes SaaS pricing less sensible. Financial planning tools are a perfect example. There will be intense periods of usage, followed by periods of much lighter usage. Paying for a full-time seat year-round doesn’t align with the value delivery pattern.Transactional Pricing: Input-Based vs. Output-BasedWhen we think about transactional pricing, it breaks down into two distinct categories: input-based and output-based.Input-based pricing is when you bill on consumption or usage that ties very closely to traditional API-based usage pricing. You charge for the number of pages processed, the amount of compute consumed, or the number of API calls executed. This model is familiar to developers and technical buyers who understand the cost structure underneath.Output-based pricing is fundamentally different. It’s much more aligned with how humans work and what they deliver. What is the output? What is the unit of work being completed? If you successfully complete a task, you bill for that unit of work or that task. This is closer to how you’d pay a human contractor or employee for deliverables rather than for effort.The benefit of transactional-based pricing, whether input-based or output-based, is that you can charge more for each unit of work you’re performing. You can scale revenue much more efficiently with the growth and expanding usage of your existing customers. As they do more, you earn more, without negotiating new contracts or adding seats.The downside is predictability. You can’t forecast revenue as confidently. There might be a quarter where your revenue is off the charts because customers heavily used your solution. Then another quarter comes where they don’t need to use it as intensively, and revenue drops. This makes it much harder to predict revenue and model your business for investors and your board.The Hybrid Model: Having Your Cake and Eating It TooWe’ve seen many companies successfully mitigate transactional pricing unpredictability by implementing hybrid models. These combine elements of fixed tiered base subscriptions with usage minimums and consumption-based billing.Here’s how it typically works. There are three tiers. For tier one, you pay X amount on a monthly or annual basis as a minimum commitment. But that subscription includes X number of credits, either input-based or output-based. If you use more than your included credits, you move into the next tier, and the next tier after that. Companies then back-bill for overage at the end of every quarter based on actual usage.This structure provides baseline revenue predictability while maintaining the ability to capture value as customer usage scales. It’s becoming increasingly popular among AI-first companies that want to balance investor expectations for predictable revenue with the reality that AI products create value through consumption.The Evolution: Aligning Price with the Value of LaborAI-first products, enabled by the evolution of technology, are making pricing models possible that align much more closely with the value of labor rather than the value of software. This is something we’re really excited about at DEFY.We’ve seen tremendous potential for companies to build very large businesses on top of this ability to target labor spend. The opportunity is massive because labor budgets dwarf software budgets in almost every organization. If you can credibly replace or augment labor with software, you can charge proportionally to the labor cost saved, not just the software value delivered.How to Choose Your Pricing Model: The Four Critical FactorsIt’s extremely important to think through the specific inputs and your unique situation when determining which pricing structure makes the most sense for your business. Every model has trade-offs. Here are the four critical factors you need to evaluate.Factor 1: Frequency of UsageHow often are customers using your tool? Is it something they use day in and day out, many times per day? If so, usage-based pricing could be really complicated and might gate adoption because users would think twice before using the solution.Slack is the perfect example. If Slack used usage-based pricing, it would be much harder for you or me to keep track of whether sending a Slack message was worth X cents. We’d mentally calculate the cost before every message, which would create massive friction and reduce usage. That’s a type of product where the frequency of use is so high that customers can’t reasonably predict their costs. A flat fee makes much more sense.Conversely, if your tool is used infrequently or sporadically, usage-based pricing makes perfect sense. Customers only pay when they get value, and they can clearly see the ROI of each interaction.Really think carefully about what the impact of pricing either on a fixed basis or usage basis will have on customer adoption and expansion. High-frequency use cases almost always favor fixed pricing to remove mental transaction costs.Factor 2: Magnitude of Cost SavingsThe second factor is understanding how much you’re saving the customer compared to the alternative, which is often human labor. If you’re delivering a 10x cost reduction compared to hiring someone to do the work, you have much more pricing power and can justify premium transactional pricing.For example, if a customer would need to pay $50 per hour for a human to complete a task, and your AI can do it for $5 worth of usage, that’s a compelling value proposition that justifies transaction-based pricing. You’re directly comparable to labor costs, and the customer can easily calculate their ROI.But if your cost savings are more marginal, or if the comparison to labor is less direct, you might struggle to justify high transaction prices. In those cases, fixed pricing might make more sense because it’s harder for customers to calculate and justify the per-use ROI.Factor 3: Integration Point in the WorkflowWhere you insert yourself into the customer’s workflow dramatically impacts your pricing model and revenue predictability.Let’s walk through an example. Imagine you’re building a customer service AI solution. If the customer wants to route all of their incoming support volume through your solution first, and you handle everything you can before escalating to humans, you have massive predictability. You can look at their historical support ticket data and have a pretty good sense of the volume you’ll process. You can plan accordingly, and they have predictability in their costs.But if they only want to insert you somewhere later in the funnel after a human has already triaged the tickets, and they only route the most complex queries to you, it’s much harder to predict volume. They’re deciding when to bring you into their funnel and what percentage of volume to send your way. Your revenue becomes dependent on their operational decisions, which introduces significant unpredictability.If you can work with customers consultatively and explain why you should be at the top of the funnel processing all volume, you can help them understand why it’s beneficial to them while also giving yourself more predictable revenue. We have a portfolio company working with logistics companies where they’re the first line of contact when anyone calls. They answer what they can in an automated way, and only route complex issues to humans.Their argument is compelling. The vast majority of customer needs when they call can be automated. By handling those, you’re reducing call wait times, which leads to lower drop-off and probably reduces lost revenue. Your human team can focus on high-value, complicated tasks instead of spending expensive human time on entry-level support tickets. They charge based on the number of minutes or calls or tickets they resolve, and the customer can predict their costs based on total incoming volume.Factor 4: Budget Type and Buyer PersonaUnderstanding whether you’re selling into a software budget or a labor budget fundamentally changes your pricing model and your competitive landscape.If you’re truly replacing labor, you’re competing against fully loaded employee costs, including salary, benefits, training, management overhead, and retention risk. That’s a much larger budget pool and allows for much higher pricing. You can justify transactional pricing based on tasks completed because you’re directly comparable to what they’d pay a human.If you’re selling into a software budget, you’re competing against other SaaS tools for limited IT spend. This typically favors fixed, predictable pricing models that IT buyers can budget for annually. The comparison is software versus software, not software versus people.The buyer persona also matters enormously. A finance or operations leader buying a labor replacement is thinking in very different terms than an IT leader buying software. The finance buyer thinks in terms of cost per unit of work, productivity gains, and headcount planning. The IT buyer thinks in terms of licenses, integrations, and annual contracts.The Pricing Strategy Question: Never Compete on Price AloneOne of the most important strategic questions early-stage companies face is how to price relative to competitors. Should you charge premium prices, match the market, or undercut competitors to gain initial traction?I’m not a huge fan of trying to enter a market and take share by underpricing your competitors. That’s a slippery slope for several reasons.First, if you win on price, competitors can also drop their prices. You haven’t built any sustainable competitive advantage. You’re in a race to the bottom where everyone’s margins compress.Second, you attract the wrong customers. Are they signing up with you because you’re the cheapest option, or because you’re the best fit for what they need? Price-sensitive customers are often the worst customers. They’ll churn as soon as someone cheaper comes along, or they’ll constantly negotiate for discounts.Third, it creates false signals of product-market fit. You might see strong initial adoption and think you’ve nailed product-market fit, when in reality you’ve just found a group of bargain hunters. When you try to raise prices or competitors match your pricing, you’ll discover that the demand was price-driven, not value-driven.That means you might have higher churn down the road, and customers might demand you drop prices further when your competitors inevitably respond. There’s a real risk to scaling your business thinking you have product-market fit when you actually don’t.I totally agree with founders who want to price at par with existing competitors rather than undercutting them. If you can’t win at comparable prices, you probably don’t have a strong enough value proposition yet. Fix the product and the positioning before trying to fix the problem with pricing discounts.Determining the Right Model for Your Specific BusinessWhether to charge for usage or flat fee or seat-based pricing requires deep customer discovery. You need to understand how customers see your product and the value it delivers.Do they see it as a labor replacement? If you’re able to do something end-to-end or truly improve the productivity of an individual, you can make a compelling ROI argument for why you should charge a premium price on a usage basis for that work.Or if they’re using it so much and so frequently that usage tracking becomes burdensome, you need to recognize that. If they’re only going to use your product for the most complicated, edge-case scenarios, then you’re actually limiting your scope and your revenue potential. You might be better off with flat pricing that encourages them to send you more volume, including the easier cases that build your data flywheel and improve your models.Understanding what your customer is optimizing for is crucial. Are they optimizing for cost predictability? For variable costs that scale with their business? For ROI on specific tasks? The answer should drive your pricing model.Your First Customers Matter More Than You ThinkBeing very careful about who your first five to ten customers are is critically important for AI-first products.You likely have to do significant work, customization, and training to make your product effective for each early customer. If they’re not willing to pay you properly, or you don’t have confidence that customer is going to stick around and provide valuable feedback, it might not be worth the effort to sign them up at all.Early customers set the precedent for your pricing, your positioning, and your product development roadmap. If you start with customers who are only willing to pay bargain-basement prices, you’ll struggle to raise prices later. If you start with customers who see you as a nice-to-have rather than a must-have, you’ll build features that serve that perception.Focus on customers who have the problem you solve most acutely, who have budget to pay for solutions, and who will be true partners in developing the product. Don’t optimize for customer count in your first ten or twenty deals. Optimize for customer quality, reference-ability, and learning.The Bottom LinePricing AI-first products is fundamentally different from pricing traditional SaaS because the value proposition is different. When you can complete tasks end-to-end or dramatically amplify human productivity, you’re competing with labor costs, not software costs. That opens up massive TAMs and justifies transactional pricing models that scale with value delivered.The key is understanding your specific situation across the four critical factors: usage frequency, cost savings magnitude, workflow integration point, and budget type. There’s no one-size-fits-all answer. SaaS pricing still makes sense for high-frequency tools where usage-based billing would create friction. Transactional pricing makes sense when you’re directly replacing human work or processing variable workloads. Hybrid models are emerging as a powerful middle ground that provides predictability while capturing scale.Whatever you choose, don’t compete on price alone. If you can’t win at market rates, you don’t have product-market fit yet. Fix that first. Your early customers and your early pricing set the trajectory for your entire business. Get them right, and everything else gets easier.The era of pure SaaS pricing is ending. The companies that figure out how to align their pricing with the labor value they deliver, rather than the software value they provide, will capture the largest TAMs and build the most valuable businesses in the AI era.* This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

So we’ve spent a ton of time over the years on SaaS talking about hiring a great VP of Sales / CRO . Not only because it really matters, but because hiring the wrong VP of Sales can set you back a year — or longer.So I thought I’d come back to the classic topic and make a list of the Top 10 Mistakes I See Founders Make When Hiring a VP of Sales:#1. Hiring a VP of Sales Who Never Really Understands Your Product During The Interview ProcessOk I know some even many will disagree, but I’m right here :). I can tell you as a pretty good investor across many leading B2B companies, I’ve never seen a VP of Sales thrive that didn’t really understand the product during the interviewing process. Never. I see so many B2B startups hire someone likeable, who can talk the talk on sales hiring and processes — but never really understands what you do. Or puts in the effort to do so. Don’t make this hire. They never invest the time after they start, either. Or they are never able to.This has almost become my #1 flag now. Way too many folks give managers a pass here that never understand the product. You gotta watch the YouTube videos. Do a demo. Listen to some Gong calls. At least get close. Or you just plain never do once you start. So many VPs of Sales disagree with me here — at least at first when I make the point. But later, they agree 😉#2. Hiring a VP of Sales With No One Lined Up to Follow ThemThis is a classic SaaStr point and post from over the years, and it turns out it’s more true today than ever. 50% of what a VP of Sales really does is recruiting. So the best VPs of Sales always have at least 2-3 great folks lined up to come with their to their next role. Just ask. Ask who those 2-3 are. And if you’re ready to extend an offer, talk to them before you do.#3. Hiring a VP of Sales That Actually Doesn’t Want to Sell Themselves AnymoreThis one has really become an issue in recent years, and the one hand I get it. Sales is hard. And it never really gets easier. So at some point in their careers, some some leaders don’t really want to sell themselves anymore. They’ll manage a team. Check the dashboards. Build process. But sell themselves? They’re sort of done. We call this Mr/Ms. Dashboards, and it’s not a new thing per se. But it’s much more common than a few years back. Because SaaS is getting to be 20+ years old.Don’t hire this person. No matter how well they can talk the talk.#4. Hiring a VP of Sales That Doesn’t Want to Go Visit Customers In PersonThis is newer, but common these days. I recently interviewed a seasoned VP of Sales that lived in the South Bay in the Bay Area. He said he wouldn’t travel all the way to SF to visit customers because it was “too far”. I get it, with traffic, it can take 90 minute. But give me a break.There are sales jobs that are 100% on Zoom. But you gotta at least visit the bigger ones. Many don’t want to do that anymore after years of working from home. Unless you sell 100% to SMBs, probably don’t make this hire. Ask.#5. Hiring a VP of Sales That Doesn’t Want to Close At Least Some Customers ThemselvesYour VP of Sales can’t carry a bag forever, at least not a full quota. But I’ve come to see that a new VP of Sales that doesn’t want to close deals themselves when they start often never really learns how to do it at all. A VP of Sales candidate that insists on closing deals themselves when they start? A great sign. One that says it doesn’t matter, that it’s all process? Maybe run.#6. Hiring a VP of Sales That Has Gotten Cyncial on Startups, Tech, and SalesSomething I didn’t use to see much, but now is pretty common. I get that everyone has a tough startup experience or two. But if you can’t get past it, if the “system is rigged” against you … well I hear you. But don’t make this hire. You need Pirates and romantics in a startup, folks whose energy drives and guides and leads the team. Not someone who sees the whole system rigged against them.#7. Hiring a VP of Sales Constantly On Social Media, Especially LinkedInI do believe some of this promotion is good. It helps with recruiting, and more. But the VPs of Sales that are posting 2-3 times a day on LinkedIn? I’ve found they really want to be influencers, advisors, etc. They don’t really want to do the tough, full-time job of VP of Sales. I know some will challenge me here. A few great posts a week on social can be good. But a few a day? Run.And yes, I know I and SaaStr post a lot on the socials 🙂 But that’s our job, folks.#8. Hiring a VP of Sales That Really Wants to Be COO, CRO, etc. And Not Really Be a VP of Sales.Don’t force someone here. If a VP of Sales is done with that role and really wants a “bigger” job where they don’t just own the new bookings number, that can have a place. But it’s not as your VP of Sales. Now a little titlle inflation IMHO isn’t the end of the world. If your VP of Sales wants to be called CRO but their real job is VP of Sales, not also owning marketing, customer success, etc. — that can be OK. As long as you’re 100% clear here. 100% clear.#9. Hiring a VP of Sales That Hasn’t Been a VP of Sales in a While But Wants to “Get Back to Sales”I get this might work in a few cases. But 95 times out of 100, don’t make this hire. VP of Sales is a tough job. Taking a short break? No problem. But going off and doing something else for a long time? I rarely see them really able to get back in the saddle again. Once in a while, yes. But understand it’s a big risk you are taking.#10. Hiring a VP of Sales That You Wouldn’t Hire If They Hadn’t Worked At ________.Ok my top flag is where we started the post — a VP of Sales that never understands what you do. But this one is close. Everyone gets blinded by that great LinkedIn, by that fancy logo on the resume. If you love Doug, but in part it’s because he worked at Datadog, or Snowflake, or Asana, or wherever, so be it. Just be honest. Would you still hire him or her if they hadn’t worked there?I ask founders to do this one exercise: block that fancy logo off from their LinkedIn. Literally, with your hand. Now, would you still hire them? If so, go for it. If not? You’re being blinded.Thanks for reading SaaStr AI: How To Sell, Scale, and Win! Subscribe for free to receive new posts and support my work. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

Brett Queener is Partner at Bonfire Ventures, a $1B AUM seed-stage fund writing $3-4M checks into application software companies. He was employee #70 at Salesforce.com, where he built go-to-market, launched the AppExchange, and helped scale the company from its earliest days. Previously, he worked at Siebel Systems (the fastest-growing software company of its era) and ran a B2B startup (SmartRecruiters) from pre-revenue to $100M ARR. He writes about the changing software industry in real-time at his Substack.He came to SaaStr Annual + AI Summit for a deep dive on AI and Product.Brett’s Top 5 Take-Aways* Your product has to deliver immediate, “What The Frack” Value Now in the Age of AI. It has to immediately do a job you couldn’t do before. * Start small, expand fast. Forget the big enterprise land. Demo with their data, put it in their hands immediately, let them feel the “holy s**t” moment with an agentic assistant—then expand. The McGillaguer-Guerrilla deal is over.* Your product must teach itself. When you’re shipping every 30 days, quarterly release webinars are dead. Build agentic assistants that tell users: “Hey, you know I can also do this? Want to try it?” The product needs a relationship with the user.* Rethink annual contracts. If agents behave like $200K employees (paid monthly, can be fired), why are we doing annual upfront payments? The renewal decision isn’t “does our software need to keep running”—it’s “is this assistant still the best person for this job?”* Fire customers who don’t get it. Some enterprise buyers want 12-month Accenture rollouts. They’re treating your agentic solution like it’s PeopleSoft in 2003. Walk away. They’ll slow you down until you die.What happens when product innovation accelerates 10x? Everything you know about building SaaS is about to change.I’ve been in enterprise software since the green screen days. Built CRM systems on Access and Visual Basic. Was employee 70 at Salesforce when we spent 60 cents of every dollar on our own data centers. Ran a startup that hit $100 million ARR (through a lot of tears, I’ll admit). Now I’m a partner at Bonfire Ventures, writing $3-4 million seed checks into application software companies doing $500K-$1M in revenue.And I’m anxious. Not the normal founder anxiety. A different kind. The kind that comes from watching the fundamental rules of software change in real-time.The Old Playbook is DeadWhen SaaStr started, the model was simple: build a SaaS version of an on-premise category winner. Ship a big product release once a year at Dreamforce or your equivalent user conference. Run the company full-throttle across the entire organization based on that annual cadence.Every blog post, every playbook, every piece of advice you’ve read about scaling SaaS? It’s all predicated on a world where your product changes once a year.But what happens when your product has to change every 30 days?Because if you’re a founder in this room right now, that’s your new reality.We’ve Never Seen This Pace BeforeI’m 55 years old. I was the first person at my company to buy a PC. I built CRM systems by carving regional manager data off IDEO drives and FedExing hard drives around the country. There was no email.I’ve watched the internet emerge. Watched SaaS kill on-premise. Watched mobile kill desktop. Watched cloud infrastructure commoditize.And I’m telling you: we have never, ever seen this pace of change in technology.Not even close.The Agentic Revolution Changes EverythingHere’s what’s different about AI agents versus every other technology shift:When you give someone an agentic assistant that actually helps them do their job for the first time, their reaction is: “Holy s**t.”Not “this is interesting.” Not “I’ll think about it.”Holy s**t.That’s a different buying motion. That’s a different renewal decision. That’s a different everything.The Three Fundamental Shifts You Need to Understand1. The Job-to-be-Done Framework Just Got ComplicatedThink about pricing for agents. Here’s my framework:* If AI helps the existing user do their current job better: Include it in the base price. Don’t charge extra.* If AI solves an additional job for the same user: Consider charging for it.* If AI does a brand new job for a new user in the org: You can charge for it.I haven’t fully wrestled this through. Nobody has. Manny (who just left Outreach) is building a CPQ platform specifically for agent pricing. That’s how nascent this is.2. Your Sales Motion Must Be “Start Small”I have a three-time successful founder in my portfolio right now. Tried to sell a full enterprise solution into a vertical with legacy, non-cloud systems. The buyer came back and said Accenture would do a 12-month rollout, starting in Ulaanbaatar before tier-one countries.I told him: “Go tell the customer they’re f*****g stupid. Tell them they don’t get it. We’re not doing 12-month rollouts.”The new motion:* Demo with sample data* Demo with their data* Put it in their hands immediatelyWhether that’s self-serve, paid trial, or assisted doesn’t matter. What matters is they start using it and feel that “holy s**t” moment.Then you expand.3. The Product Must Teach ItselfWhen you’re shipping major updates every 30 days, the old “quarterly release webinar” model is dead.The future? Your agentic assistant says: “Hey Brett, you’re using me for X, Y, and Z. You know I can also do A? Would you like to try it?”The product has to have a relationship with the user that helps them discover new capabilities as they ship.What This Means for Your Company StructureIf you’re shipping quarterly in 2025, you’re already behind.But here’s what nobody’s talking about: when you move to continuous shipping, your entire go-to-market org has to change with it.I used to invest in SaaS companies that said “we ship daily.” I’d tell them: “I promise you’ll move to quarterly releases. Once you build a sales team, success team, and customer team, you’ll need coordination.”That’s still true. But the coordination has to happen at 10x speed.The challenge: How do you ensure your customer-facing teams know about new capabilities fast enough to drive adoption before the next release ships?Most companies will fail at this. Their product will race ahead while their GTM team sells last quarter’s product.The Renewal Decision is Fundamentally DifferentWhen I hire a $200K assistant, do I pay them $200K upfront? No. I pay them monthly. Can I fire them? Yes (with some severance).So why are we still doing annual upfront payments for agentic solutions that behave like employees, not software?The renewal decision for an agent isn’t: “Does our Salesforce org need to keep running?”It’s: “Is this assistant still the best person for this job compared to other assistants I could hire?”That changes everything about retention.Your customers need to understand all your capabilities in a way they can consume. Not once a quarter. Not through a CS quarterly business review. Continuously. In the product.The Uncomfortable Truth About Enterprise BuyersSome enterprise buyers just suck.They don’t understand this shift. They’re still thinking in 12-month implementation cycles. They’re routing everything through Accenture.And here’s what I’m telling my founders: You might need to fire those customers.Not literally (usually). But you need to be willing to walk away from buyers who want to treat your agentic solution like it’s PeopleSoft in 2003.Because if you let them slow you down to their pace, you’ll die. Some competitor who’s shipping every 30 days will eat your lunch.What To Do Right NowIf you’re a founder reading this, here’s your checklist:Immediate (This Week):* Audit your product release cycle. If it’s longer than 30 days, you have a problem.* Look at your last 5 deals. How long from demo to “hands on keyboard”? If it’s more than 2 weeks, you have a problem.* Check your pricing. Are you charging for AI capabilities that make the core job better? Stop.This Quarter:* Build the mechanism for your product to teach users about new capabilities* Restructure your sales comp to reward “start small, expand fast” not “land the biggest deal possible”* Identify which enterprise buyers “get it” and which don’t. Plan accordingly.This Year:* Question everything about annual contracts and upfront payments* Build the muscle to ship major capabilities monthly, not quarterly* Accept that your GTM playbook from 2019 is deadThe Bottom LineWe’re in the change economy now. The rules that governed SaaS for the last 20 years are being rewritten in real-time.Product innovation is 10x faster. Buyer expectations are fundamentally different. The very nature of what “software” means is shifting from tools to teammates.Most companies won’t make the transition. They’ll keep running the old playbook, shipping quarterly, doing 12-month enterprise rollouts, charging annual upfront.And they’ll die.The companies that survive will be the ones who embrace this pace of change. Who ship every 30 days. Who start small and expand fast. Who let their product teach itself to users.It’s uncomfortable. It’s anxiety-inducing. We’ve never seen anything like this before.But that’s exactly why it’s the biggest opportunity in software since the shift from on-premise to cloud.Are you moving fast enough?Thanks for reading SaaStr AI: How To Sell, Scale, and Win! Subscribe for free to receive new posts and support my work. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

A SaaStr Annual + AI Summit conversation with Jesse Zhang, CEO of Decagon, and Sarah Ittelson, Partner at AccelDecagon Today: The Numbers Behind the HypeFounded in late 2023—just months after GPT-4’s release—Decagon has become one of the fastest-growing AI companies in history. The company builds AI customer service agents for large enterprises, automating conversations that previously required human support teams.The Growth Trajectory:* Founded: Late 2023* Time to Eight Figures ARR: ~18 months* Team Size: ~100 people (and scaling rapidly)* Location: 100% in-person team* Customers: Major enterprises including Hertz, Chime, and other leading brands* Typical Customer ROI: $800K in savings for every $250K spent* Market Position: Recognized as the leading Gen-AI native solution in customer service automationWhat Makes This Growth Unprecedented:Even by venture standards, this is exceptional. Sarah Ittelson, the Accel partner who led their Series A investment, has been part of the hyper-growth phases at Uber, Uber Eats, and Fair. Her assessment? “This current moment and the scaling that’s possible within these AI companies is unparalleled to even those hyper-growth moments of before.”When Accel invested at the Series A, Decagon was targeting seven figures. By the time Jesse and Sarah took the stage at SaaStr to share their playbook, they’d already blown past eight figures. The headline had to be updated mid-flight.This isn’t a story about getting lucky in a hot market. It’s a masterclass in intentional decision-making, relentless customer focus, and building a machine that compounds growth. Let’s unpack exactly how they did it.The Market Selection Framework: Why Customer Service WonHere’s the reality most founders miss: your growth rate is mostly determined by which market you’re in.Jesse and his co-founder didn’t just pick customer service because it seemed like a good idea. They ran a rigorous discovery process—talking to roughly 100 potential customers over the course of a month. Every day packed with customer conversations. Every night cranking out product to show the next day.What made customer service the winner?* Clear, measurable ROI: Companies could point to specific dollar savings. Spend $250K, save $800K in human support costs. That’s not a pitch—that’s math.* Massive TAM: Customer service is one of those rare markets where the surface area is enormous. Every large company has support teams. Every user interaction is a potential automation opportunity.* Buyer urgency: Unlike many SaaS categories, companies were willing to move off schedule to adopt AI solutions. The business case was too compelling to wait for the next budget cycle.The key insight: they weren’t looking for a market where AI could work. They were looking for a market where companies were already bleeding money on a problem that AI could solve today.The Customer Discovery Process: 100 Conversations Before a Single Line of CodeLet’s get tactical about how Decagon approached early customer discovery, because this is where most founders either win or waste months.The Daily Cadence:* Pack every day with as many customer conversations as possible* Extract commitments: “If we built this, how much would it be worth to you?”* Build at night based on what you learned during the day* Show it to customers the next day* Iterate ruthlesslyWhat They Were Really Listening For:Not excitement. Not validation. They were listening for willingness to pay.Here’s Jesse’s framework: You could talk to the wrong person who feeds you useless feedback. Or you could talk to someone who’s super excited about your pitch, but when you get to pricing, there’s nothing there. True discovery means being aggressive about understanding what customers actually value and what they’ll actually pay for.The Anti-Pattern They Avoided:Jesse’s previous company was a consumer startup (eventually acquired by Niantic). The biggest lesson? They spent too much time sitting by themselves thinking about good ideas, building them, launching, and getting zero traction. That’s the burnout loop.With Decagon, they flipped the script: talk first, build second, validate constantly.Staying Close to Customers at ScaleHere’s where it gets interesting. Most companies do customer discovery well early, then lose touch as they scale. Decagon hasn’t.How they maintain customer intimacy:* Weekly touchpoints with every customer (they work only with larger customers, which makes this feasible)* Proactive outreach—don’t wait for customers to complain* The entire team stays involved in go-to-market, not just salesThe principle: customers might not proactively tell you what’s missing. You have to pull it out of them. You have to understand what’s next on the horizon before they fully articulate it.Building the Team: Why In-Person and Intelligence MatterWhen Sarah Iden first met with Decagon, Jesse was closing employee number five. Even at that stage, the focus on talent was exceptional.The Decagon Team Philosophy:They’re essentially 100% in-person. That’s not an accident—it’s a forcing function for the culture they wanted to build.What They Screen For:* Raw intelligence and adaptability: At the early stage, there’s no structure. No playbooks. No one hands you a to-do list. You need people who can figure it out.* Commitment level: This is the non-negotiable with co-founders and early employees. There’s a wide range in how committed people are to being a founder or early-stage employee. If it’s not aligned, nothing else matters.* Working style alignment: Are people working at the same pace? Putting in similar time? Gearing toward the same goals?The Interview Process:They ask about past experiences: “Tell me about a project where you invested the most time in your life.” They look at favorite projects, how hard people worked, the pace they maintained. They get back-channel references. By the time someone reaches the final interview with Jesse, they’ve already opted into the in-person culture and the intensity.There’s no hard sell needed. The screening has done its job.The Co-Founder Equation: Finding Your MatchJesse’s co-founder relationship formed through A16Z (who backed his previous company) and mutual friends. But the chemistry wasn’t luck—it was alignment on fundamentals.What Must Align:* Similar life stage* Level of commitment (the absolute non-negotiable)* Working style and pace* Type of company you want to build* Amount of time you’re willing to investWhat Can Differ:* Specific skill sets* Personality traits (to a degree)The Best Way to Find Out:Build something together. Do a trial. If you’re considering co-founding, you probably have the time to experiment. This is unique to co-founder dating—you can’t get this kind of trial when hiring someone with a full-time job.The trial will surface everything: working style, commitment level, how you handle disagreement, pace of execution. Everything else is just conversation.The Execution Culture: How to Maintain Intensity at 100 PeopleIn the early days, execution culture is easy. Everyone’s in one room. You’re talking constantly. Alignment is organic.At 100 people, it requires systems.Decagon’s North Star Approach:Everyone needs to know what they’re working toward. If you’re in sales or engineering, you need clear goals aligned in the same direction. If you’re not working toward one of the important goals, you’re probably not working on something important.The Metrics That Matter:For a SaaS company, it’s simple:* ARR* Customer count* Customer qualityThese are measurable. They become the north star everyone marches toward.The Celebration Cadence:They celebrate wins constantly:* Customer gives great feedback* Customer goes live* Deal closes* Customer renews for two yearsEvery win gets shared. The team knows what they’re working toward and they see the impact in real-time.The Challenge Ahead:Jesse’s candid about this: they’re still figuring out how to maintain this culture as they scale. Right now they’re about 100 people and growing quickly. At some point, you can’t maintain culture just by talking. You need systems. They’re actively building those now.The AI-Native SaaS Motion: Why Everything’s DifferentHere’s what’s fundamentally changed about selling SaaS in the AI era:1. ROI Must Be Provable and FastYou can’t hand-wave anymore. Customers expect to test and see real ROI. For Decagon, that means:* Quantifiable conversation resolution rates* Measurable cost savings* Improved customer satisfaction scoresIf both cost savings and CSAT are going up, the business case sells itself. They’re saving money AND customers are happier. That drives future revenue and retention.2. Deployment Can Be IncrementalYou don’t have to deploy all at once. Pick one surface area. Deploy to 10% of your customer base. Get live in production. If it works, ramp up.This makes buying cycles shorter and happens more frequently because you’re not waiting for massive projects with perfect test cases before commercial discussions begin.3. The Technology Layer Keeps ChangingModels are improving constantly. New techniques emerge. You have to build with the assumption that everything underneath will change.How Decagon Handles This:* Build for model flexibility—instant evaluation and swapping of new models* Own the application layer (the final product customers buy)* Focus on solving the business problem, not showcasing the technologyThe business problem doesn’t change even as technology evolves. That’s the power of building at the application layer. You’re evaluated on solving the problem, and as technology improves, your product improves automatically.Breaking Through the Noise: Product Marketing in a Hyped CategoryEvery company is being told to have their “AI narrative.” So how does Decagon cut through?Message #1: Gen-AI NativeCustomer service has a long history of automation. The natural question is: “How are you different from existing solutions that are now adding Gen-AI?”Decagon’s answer: They built from the ground up for Gen-AI. The entire paradigm of how you build these agents is different. You can do fundamentally more when you’re not retrofitting Gen-AI onto legacy architecture.Message #2: Real ROI with Real CustomersAI is hyped. But Decagon has case studies with companies like Hertz and Chime showing:* Specific dollar savings* Support team repurposing* Higher NPS and retentionThe message isn’t “AI is cool.” It’s “AI actually works here, and here’s the proof.”Goal Setting: Maintaining Healthy DissatisfactionDespite the eight-figure run rate and explosive growth, there’s no back-patting at Decagon.Why:The market is massive. They’ve touched a fraction of a fraction of it. No one feels like they’ve achieved the goal yet.The final vision: working with the biggest companies out there, deployed across entire surface areas, handling the entirety of customer interactions.When your goal is that far in the future, it’s easy to stay hungry. There’s always so much more to do.The Meta-Lesson: Market Selection Trumps EverythingJesse’s closing advice deserves its own section because it’s the thread running through everything:Pick the right market.Within SaaS, if you can find a market with the right properties—clear ROI, measurable outcomes, large TAM, buyer urgency—it makes everyone’s life easier. The structured nature of SaaS (concrete customers, measurable revenue) compounds when you’re in the right market.His previous consumer company could grow faster but had way less structure, higher variance, and no single customers paying meaningful money that you could talk to deeply.The lesson: founder-market fit matters, but so does market-outcome fit. Choose a market where the fundamentals work in your favor.The Q&A: How to Break Into Enterprise CustomersQuestion from audience: “How do you actually get those first customer conversations? How do you connect with companies like Rippling or Notion?”Jesse’s Answer:Find companies known for working with startups. Rippling is one. Then find the shortest path in:* Maybe your friend knows someone* Maybe your VC knows someone* Maybe it’s two hops awayDo it enough and you’ll get through. Maybe the first person isn’t right, but their coworker has a use case. It’s about finding the shortest path and doing it again and again.When they were starting, they aggressively lined up conversations with literally anyone who would talk to them. That’s just what you have to do.The Bottom LineDecagon’s story isn’t about luck. It’s about:* Choosing a market where ROI is measurable and urgent* Talking to 100 customers before writing code* Building a team that reflects founder values and commitment* Maintaining customer intimacy even as you scale* Creating an execution culture with clear north stars* Building for flexibility as AI technology evolves* Staying hungry no matter how fast you growZero to eight figures in 18 months is extraordinary. But the playbook is surprisingly replicable: pick the right market, talk to customers obsessively, hire for intensity and intelligence, execute relentlessly, and never stop pushing forward.The AI-native SaaS wave is real. But winners won’t be determined by who has the best models or the most hype. Winners will be determined by who executes this playbook most effectively.Decagon is showing us how it’s done.Thanks for reading SaaStr AI: How To Sell, Scale, and Win! Subscribe for free to receive new posts and support my work. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

The State of AI + Software: Where It’s Going - FastThis deep dive is from Jason Lemkin at the LIVE AI Workshop Wednesday. Sign up here for the next one.I was talking to a founder recently who’s running at $50 million ARR. Classic SaaS guy turned AI guy. And he’s going to scale from $50M to $100M with just five sales reps and a team of AI agents.In the old days, at $50 million, you’d probably have at least 100 sales reps. Why? Because to get from $50M to $100M, you need $50M in net new bookings. At $500K net per rep (which is pretty good when you factor in scaling, turnover, and ramp time), you’d need 100 bodies. Minimum.This founder? Five human reps. Plus AI.It’s not that he doesn’t need sales reps. It’s not even that he’s not selling. He’s actually doing classic B2B SaaS sales. He’s just doing it with dramatically fewer humans, and for the purposes of this discussion, he’s doing it so much more efficiently. And he has 10,000+ inbound leads a month flowing through this system.This is where we are right now. And if you’re not paying attention, you’re already behind.Everything Changed For SaaStr Itself in the Last 180 DaysAt the end of Q1 this year, we had zero AI agents in production at SaaStr. Nothing. Nada. We were thinking about it, but we hadn’t deployed anything.Fast forward to today, and we’ve got:* Almost 20 AI agents running in production* Four different AI SDRs deployed and actively working leads* Salesforce Agent Force rolled out (just started yesterday, so we need a bit more time before sharing all the data)* An AI BDR from Qualified handling inbound qualification* A slew of specialized agents for support, research, and operationsIt is so much different even on our little team than it was even 100 days ago. And it’s going to keep changing at this pace.I’ll be honest: we were probably a little behind the curve at the start of the year. Now, we’re kind of at the bleeding edge. And we want to drag everybody along with us because everything’s changing.The models are changing. The tools are changing. Things that didn’t work last year can work really well now.Everyone complained about how crummy AI SDRs were last year—and there still are a lot of issues—but now we know how to train them. Now we know how to iterate with them. Now we know how to make them work.There’s so much more coming, and marketing in some ways is even further behind sales. But it won’t be for long. This whole space is going to radically change in the next 12 months.More on our AI Agents here.The Single Most Important Thing You Need to Do to Stay RelevantHere’s my advice, and I mean this with every fiber of my being:If you feel behind in AI and go-to-market, the answer is simple: be part of a deployment.Not just “buy a tool and forget about it.” That teaches you absolutely nothing.Go buy any tool. Go buy any tool that comes and talks at events. Go buy any leading tool in the space. At some level, it doesn’t even matter which one you start with.But here’s the critical part: Don’t just buy it. Be part of the deployment.* Train it yourself* Be part of the onboarding* Be part of the errors and the issues* Be part of the daily iterations you have to do to get it workingDon’t just set and forget. You will learn nothing. You will learn absolutely nothing.Be part of a deployment to see how it really works. Otherwise, you’ll never really learn.You don’t have to become a vibe coder like I have. You don’t even need to play with the underlying tech. What you really need to do is get your hands dirty with actual implementation.The Iconiq Growth Report: 10 Things That Show How Much the World Is ChangingIconiq Growth—one of the top late-stage B2B growth funds forever, from the early days when they managed Mark Zuckerberg’s family money to being in so many leaders from Canva to Anthropic and beyond—just put out an incredible report.https://www.saastr.com/iconiqs-state-of-software-in-2025-much-smaller-teams-55-of-gtm-is-now-in-post-sales-2025-is-crushing-2024-for-funding/It’s almost 100 pages. There’s a lot of detail in there about public companies and nuances that a lot of you don’t really care about. But I picked out what I think are the 10 most relevant things to SaaS folks for B2B that show just how much the world is changing.A lot of folks are behind here. A lot of folks. And it’s okay—it’s not necessarily fatal. We’re not all being killed by ChatGPT and Anthropic. But the world is changing. This data shows you just how it’s changing and how we all have to adapt.1. AI Native Companies Are Burning Less (Way Less)This first data point explains a lot of things that are really confusing on social media. People say, “Oh, these AI native companies, they’re burning all this cash to acquire customers with free tiers and consumption models.”Actually? Not true.AI native companies burn substantially less than classic SaaS companies did at the same stage. Why? Because they don’t need to hire armies of people to scale. The unit economics work differently when AI is doing a significant portion of the work.2. The Sales Team Size CliffWe’re seeing companies hit $10M, $25M, even $50M ARR with radically smaller sales teams than the old playbook required.The old rule of thumb: you needed roughly one sales rep per $500K-$1M in ARR, depending on your ACV and sales cycle.The new reality: AI-native companies are operating with 20-30% of the headcount at equivalent revenue levels.That $50M company I mentioned at the beginning? That’s not an outlier. That’s the new normal for AI-native companies.3. CAC Payback Periods Are CompressingCustomer Acquisition Cost (CAC) payback periods are getting shorter in AI-native companies. Not because they’re spending less on acquisition (though some are), but because they’re able to onboard, activate, and expand customers faster with AI-powered customer success and product experiences.We’re seeing companies hit CAC payback in 6-8 months that would have taken 12-18 months in the classic SaaS playbook.4. Net Revenue Retention Is Under Pressure EverywhereHere’s the uncomfortable truth: Net Revenue Retention (NRR) is declining across the board.Even the best SaaS companies are seeing NRR pressure. Why?* Competition is fiercer than ever* Customers are more price-sensitive post-2021* AI is enabling customers to do more with less, reducing expansion* New AI-native alternatives are constantly emergingThe days of 130%+ NRR being table stakes? Those are largely over. Now, 120% is excellent. 115% is solid. And anything above 110% in the current environment is worth celebrating.5. The Growth Multiple Advantage of AI CompaniesHere’s where it gets really interesting: AI companies are trading at higher multiples than pure SaaS companies at equivalent growth rates.Why? Because investors believe:* The TAM is larger (AI can address bigger markets)* The margin structure is better long-term* The defensibility is stronger (data moats, model improvements)* The growth trajectory is more durableA SaaS company growing 100% YoY might trade at 10-15x ARR. An AI company growing 100% YoY? We’re seeing 20-30x+ ARR in some cases.6. Time to $10M ARR Is AcceleratingAI-native companies are hitting $10M ARR significantly faster than classic SaaS companies did.Classic SaaS: 3-5 years from founding to $10M ARR was typical for fast-growing companies.AI-native: We’re seeing companies hit $10M ARR in 18-24 months.Why? Product-led growth combined with AI efficiency. Lower go-to-market costs. Faster time to value. Viral distribution.7. The Disappearance of the Middle-Market Private Equity ExitThis one is painful but important: The private equity exit path for $20M-$50M ARR SaaS companies has essentially disappeared.In the 2010s and early 2020s, if you had a solid SaaS company growing 30-50%, your burn rate was low, and your NRR was high, maybe VCs wouldn’t touch you, but a private equity firm would come in and buy you for 6x to sometimes 10x revenue.You’d get to $20M revenue growing 50-60% and cash flow neutral. No VC was going to fund you there, but a PE firm might buy you for $150M or even $200M.I don’t see any of those deals anymore. They’ve disappeared.Why? Because private equity isn’t immune to the fact that:* Public SaaS companies have seen growth decay* NRR isn’t as sticky as it used to be* New AI vendors are putting the old guys at riskSo PE has disappeared from this segment, and VCs have flocked to hyper-growth AI companies.This may or may not change, but at least don’t live in a dream world where you think the next round is going to come because growth is “pretty good.” It won’t.8. If You Have Any Doubt About Fundability, Check the DataWe literally added a new benchmarking tool at SaaStr.ai. It’s our third tool there. Go to saastr.ai, click “AI VC” at the top.Upload your latest investor update or board deck or VC pitch. We will tell you the exact odds you get funded. The exact odds.And if you don’t like it, don’t shoot the messenger. It’s based on all the recent data from 5,000+ rounds.I talk to way too many founders who think they’re going to get funded in B2B, and they’re not. So just at least find out. We built the tool. It’s benchmarking. It’s cool.9. AI Is Creating a New Category of “Efficiency Giants”We’re seeing the emergence of what I call “efficiency giants”: companies that can reach $100M+ ARR with teams that would have been laughably small in the SaaS 1.0 era.These companies have:* Sub-50 person teams at $50M+ ARR* Gross margins in the 85-90% range* CAC payback under 6 months* Growth rates still above 100% YoYThese aren’t just outliers. This is becoming a category.10. The AI-Native Stack Is Real and GrowingCompanies are increasingly building on what I call the “AI-native stack”: infrastructure, tooling, and platforms designed specifically for AI workloads and AI-powered applications.This includes:* Vector databases (Pinecone, Weaviate)* LLM orchestration layers (LangChain, LlamaIndex)* AI observability and monitoring tools* Model fine-tuning and deployment platformsThe total spend on AI-native infrastructure is growing 300%+ YoY, creating a massive new category of B2B software.The Hard Truth About Introducing AI SDRs to Your TeamSomeone asked me a great question in the live AMA: “How do you start to thoughtfully introduce AI SDRs to the sales team without scaring off the existing SDRs?”I was literally dealing with this yesterday. I was at the company all-hands for a company I invested in called MangoMint (SaaS + AI Agents for salons, spas, and doctor’s offices). Great company, great CEO, great culture. People were just so excited at their all-hands.The CEO asked just one favor of me, “Just do me one favor: don’t scare the SDRs.” He said, “Don’t scare them with all this stuff.”I said, “Okay, I won’t.”But actually, I’d already had lunch before the presentation. I sat down at tables with all the SDRs. I’d already had the conversation with them.They understood that the best have as important a role as ever, and the others don’t.The VP of sales sat down with me afterward and gave the same message. She said, “We’re aggressively all-in on these tools, and we’re growing. We’re growing triple digits at 8 figures in revenue. This is the future. We have to embrace it.”So look, here’s the hard truth: If your SDRs are going to quit because of AI, they’re going to quit anyway.And the tenure of an SDR is very short to begin with. Just be clear: we need everybody who can perform.It was funny. We were sitting there at lunch with the SDRs, and one of them was one of the best ones on Marshall’s team. I’m sitting next to her, and she’s already got her laptop up. She’s scanning a salon that she’s doing outbound to during lunch. She’s literally doing true outbound research during lunch.She’s got a job forever.The SDRs that just want to run random emails and never learn the product? They don’t.I think we’re past the point where you have to worry about scaring the team. Be kind. Make sure everyone who crushes it knows that everyone who is A-tier on your team has a role. That hasn’t changed.It’s even harder to find A-tier folks now. But if folks don’t want to adapt, they won’t have a future. And I think you’re better off being straightforward with them.Just do it.Here’s the other thing: As soon as you actually roll out the tool, someone’s going to quit anyway.This is the story with every company I’ve invested in. Even at SaaStr, on our little team, we rolled out a tool called Momentum (or it might have been from Attention—they’re great too).Even on our little team, the day we rolled it out, someone on our sales team quit. The day we rolled it out.Why? Because now all his actions, all his data were going to show up in real-time in a report and in Salesforce. You couldn’t hide. You couldn’t pretend you did five calls this day when you didn’t. You couldn’t pretend you did the work when you were out at this tool.Somebody quits. And honestly? It’s all for the best. The truth is, it’s all for the best.We’ve got to get out of kumbaya land because this is a new high-growth world. You’ve got to be a part of it.The Real Cost of Implementing AI: What You Actually Need to BudgetAnother great question: “How much should we be estimating for cost to invest per year in AI to get started?”This is better than it sounds, and the answer is going to surprise you.Let me give you the SaaStr example. We’re tiny. We spend maybe $10,000 a year on Salesforce. Maybe a little more. And the same or a bit more on Marketo, which we essentially use as a CRM for our community (Salesforce is our CRM for sponsors and ‘sales’).I’ve been a Salesforce customer for 20 years. First piece of SaaS I ever purchased for real. I never paid for anything before Salesforce.But here’s the thing: we spent maybe $10K on Salesforce notionally.We spent $500,000 on our AI agents across these 21 agents. $500,000.Think about that ratio for a minute. There are so many learnings in that ratio.It’s why Salesforce has to own Agent Force. They have to win. It’s why HubSpot has to win here.But we are spending far more on agents than we are on CRM. That’s a meta thing to think about.The Real Cost Structure of AI ToolsThe second thing to think about: these apps that need to be trained with forward-deployed engineers and hands-on work? I don’t think very many of them are less than $30,000 or $50,000 a year.A lot of them actually try to have a price point that’s approaching $100K—like $60K a year base, and then $20K or $30K for the forward-deployed engineer and training.Sometimes they include it, sometimes they charge for it. Whether they do or don’t, it ain’t free to do it right.For me, I would certainly subsidize it if I were a vendor, but it costs that amount. Having a really great technical resource help you train your AI for a couple of weeks? That’s not free. It’s pretty damn expensive.And this relates to the meta issue: You’ve got to invest the time.There’s a reason these apps cost $50K to $100K. They are a lot of money. Be wary of super cheap apps. Be wary.It’s not that we’re not getting there. It’s not that everything isn’t going to get better. But you can’t cut corners on training.If instead of $50K a year, you’re buying a solution that’s $500 a month or $50 a month, I’m not saying it’s not going to work, but you’re going to have to do even more training. You better take on a massive amount of burden to train this app because you’re not getting the benefit of the forward-deployed engineers and all the training, tuning, warming, and everything the vendor does.There’s no shortcut here. There is no shortcut.It is an existential issue. This will change over time, but right now, the agents can’t train the agents themselves the way the best humans can.It’s tough to get away with less than $30K-$50K for any of these tools when a month of training is required—or at least several weeks.What This All Means for YouLook, I get it. This is overwhelming. The pace of change is brutal. Every time you think you’ve figured something out, the models get better, the tools evolve, and the playbook shifts.But here’s what you need to take away from all of this:1. The efficiency advantage is real. AI-native companies are operating with dramatically lower costs and higher margins than classic SaaS ever could. If you’re not thinking about how to capture this advantage, you’re leaving money—and competitive positioning—on the table.2. You can’t learn by observing. You have to deploy. You have to train. You have to iterate. Reading blog posts and attending webinars isn’t going to cut it. Get in the arena.3. The funding environment has bifurcated. Hyper-growth AI companies are getting funded at eye-watering valuations. Solid but unspectacular SaaS companies are getting passed over. Know which category you’re in and plan accordingly.4. Your team structure needs to change. The old GTM playbooks are dying. You don’t need 100 SDRs to scale to $100M anymore. But you do need the right people who can work alongside AI, train it, and amplify their effectiveness with it.5. Budget for this properly. You’re going to spend more on AI agents than you think. And that’s okay—the ROI is there. But don’t go into this thinking you’re going to deploy enterprise-grade AI solutions for $500/month. You’re not.6. Speed matters more than ever. The companies that figure this out in the next 12 months are going to have a massive, compounding advantage. The companies that wait are going to find themselves 18-24 months behind, and in this market, that might be fatal.The Bottom LineWe’re in the early innings of the biggest shift in B2B software since the move from on-premise to cloud. Maybe bigger.The companies that win are going to be the ones that:* Deploy aggressively* Learn constantly* Adapt their teams and processes* Invest in the right tools and training* Move fast without breaking (too many) thingsThe companies that lose are going to be the ones that watch from the sidelines, wait for “best practices” to emerge, and try to protect the status quo.Don’t be the second group.Get in. Deploy something. Train it. Break it. Fix it. Learn from it.That’s the only way forward.And if you want to learn more about how we’re doing all of this in practice, come to SaaStr AI in London on December 1-2. We’ll have speakers from OpenAI, Whiz, Clay, Intercom, Fin, and all your favorite B2B + AI companies.We’ll do live sessions. We’ll share real data. We’ll show you what’s working and what’s not.Because this is changing every quarter, and we’re all learning together.Let’s go.Want to see if your company is fundable in today’s market? Check out our AI VC benchmarking tool at saastr.ai. Upload your deck and get your exact odds. It’s free. It’s data-driven. And it might save you months of wasted pitching.Thanks for reading SaaStr AI: How To Sell, Scale, and Win! Subscribe for free to receive new posts and support my work. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

A deep dive into the playbook, lessons learned, and brutal truths about deploying 20+ AI agents into production — that actually work.During Dreamforce Jason (CEO) and Amelia (Chief AI Officer) at SaaStr dropped by Qualified’s office for a deep dive on its AI Agents and top learnings with Kraig Swensrud (Founder & CEO, Qualified)The Bottom Line Up FrontIf you’re a CMO, CRO, or founder and you haven’t deployed at least one AI agent by Halloween 2025, you’re already behind. Not “might fall behind” — you’re already behind. “Every VC I know where a startup hasn’t made the jump yet has given up hope on that company. That’s not hyperbole. That’s the market reality.” per Jason.But here’s what under-discussed training is more important than picking the perfect vendor. We started 2025 with zero AI agents at SaaStr. Now we have 20 in production. The secret wasn’t finding magical tools — it was investing 30 days of deep training upfront, then maintaining an hour every single day.Here’s exactly how we did it, what worked, what failed, and the framework you need to deploy agents that actually drive revenue.The New Budget Reality: Why Traditional SaaS Playbooks Are DeadLet’s start with the uncomfortable truth about 2025 budgets:Traditional SaaS budgets are frozen. CEOs are going around the table telling every functional head to cut 20-30% of their SaaS apps. Half of incremental budgets are going to price increases — Salesforce raised prices 8% this year, others 6-7%. When your IT budget is growing 6% but your core vendors are raising prices 7-8%, where’s the room for another business process workflow app?But AI budgets are exploding. Business software is growing faster than it has ever before — if you tap into AI budget. It’s the only incremental budget most companies have. Nobody is putting more money into old SaaS software. It’s all going into AI.This creates a tale of two cities:* Classic SaaS is geriatric ... but ...* B2B software powered by AI is explodingIf you’re selling the way you sold in 2021, with the 2019 Marketo playbook, there’s no budget for you. The playbook doesn’t work. But tell that to anyone on fire with AI — everything works. Outbound works. Events work. Meeting with customers works. If anyone wants to buy your product, it all works.The Vendor Selection Myth: Training Trumps EverythingHere’s the question Jason gets constantly on LinkedIn: “What’s better — Replit or Lovable? Which AI SDR platform is best? What AI tool is best for RevOps?”Wrong question.Here’s what Jason learned after deploying 20 agents: Pick a leader and go deep. Training is more important than vendor selection.There are very few agentic products with any level of complexity where you can just flip a switch. Take Qualified — it set up seven appointments for us on its own. Sounds great, right? Put it in a case study. But it wouldn’t have happened if we put zero minutes of training into it.The New Deployment RealityRemember how people bought enterprise software before AI? You’d buy it from a sales rep. You might hire Accenture or Blue Wolf to deploy it over a year. You’d hope and pray it worked like the sales rep told you.No one will tolerate that in AI this year. Time to value often has to happen before you even get a contract signed.Jason is 100 days into Replit but probably 200 hours expert now to make it great. That’s a sea change in how we use software. Companies are used to buying from a people person and scaling up over time with humans doing all the work. Those days are behind us.The Training Investment FrameworkFor any agent you deploy:* Commit 30 days upfront — train it every single day* Then commit weekly — every week after that* Budget 1 hour daily ongoing — this is your new normalMost people don’t know how to do this. But if you don’t train it, you won’t get ROI. Period.The SaaStr Agent Journey: From 0 to 20 in 10 MonthsAgent #1: Jason AI (The Horizontal Foundation) aka “Digital Jason”SaaStr started with Delphi, building a clone of Jason — “Jason AI.“ It ingested all of SaaStr content (20 million words, 1,000 YouTube videos, all tweets, all LinkedIn posts). It actually broke the ingestion engine at first, but once it worked, it was magical.Key insight: SaaStr started with a horizontal agent that could do a little bit of everything. No complex workflows. Just digital Jason answering founder questions. Training was broad, but not especially deep. Perfect for a general agent, answering general questions.What happened: People started using it for things we didn’t envision. They used it for support questions about attending our events, sponsor questions, booth logistics. Classic issue with AI agents — if it’s good, people will start using it for everything. If they trust it, usage expands.Then it started to hallucinate on event-specific questions. So the team uploaded the prospectus. With that additional context and training, it got good. Maybe 20% as good as Qualified, but 20% is infinitely better than zero.Before: Pre-Fin Intercom with week-long response timesAfter: Instant answers, 90% accuracy on customer questionsThe Stair-Step Strategy That Actually WorkedHere’s the deployment sequence that worked for SaaStr:Step 1: Horizontal Agent (Delphi/Jason AI)* Easiest to deploy* Least training required* Built confidence* Got an early winStep 2: Outbound SDR (Artisan)* Lower risk than inbound* SaaStr’s SDRs wouldn’t do outbound anyway* Better to send a better email via agent than mediocre human* Less risk than trusting agents with precious inbound leadsStep 3: Inbound Qualification (Qualified)* Higher complexity* More critical to get right* But SaaStr’s qualification process was nearly broken* Low bar made deployment easierStep 4-20: Specialized Agents* Email campaigns* Website traffic deanonymization* Follow-up sequences* Support ticketing* AI speaker content review* AI VCWhy Stair-Stepping MattersIf you fail with your first agentic thing, you won’t know what went wrong. You’ll get frustrated. You won’t build the confidence to keep going.SaaStr accidentally stair-stepped their deployment, and it’s why they succeeded. They gained confidence with each win. By the time they deployed Qualified for inbound (higher stakes), the team knew how to train agents properly.The framework:* Use horizontal agents where bar is low, where there is nothing already and you have some strong content to train it on. But it doesn’t need to be a vast amount of content. It just has to be useful and relevant.* Use specialized agents at first for areas with zero human coverage. Because the bar is low for success.* Save the hardest, most critical deployments for when you have 2-3 agents under your belt.The “Layup Roles”: Where to Start Right NowThe biggest question: where do I actually start?The layup roles are where there’s a trustworthy tool that exists, but work is just not getting done in your org.There are lots of reasons work doesn’t get done:* Tough to hire in that area* Tough geography for hiring* Not your DNA (product-centric companies suck at sales, sales-centric companies suck at product)* High turnover* Junior team members who don’t know your product/market/competitorsDon’t try to replace stuff that’s working. Look at where there’s literally no one doing the work.The Proven Starting Points#1: Customer Support (The Easiest Layup)Go try your own support as a founder or executive right now. Nine times out of ten, you’ll cry at how bad it is. Submit a P1 issue. See if you get “We’ll get back to you in a week” or “I don’t understand the problem.”That’s where to start. The bar is so low, your agent doesn’t have to be amazing. It just has to be better than terrible.We saw this at SaaStr: Went from Intercom (week response times) to instant accurate answers in 90% of cases. The other 10%? The agent knew to escalate.#2: Outbound SDR/BDRFor us, this was lower hanging fruit than inbound. Why? Our SDRs couldn’t (or wouldn’t) do outbound consistently. The bar: send better emails than mediocre humans were sending — or in our case, not sending.Less risky than inbound because you’re not trusting agents with precious inbound leads. You’re just trying to generate pipeline that wasn’t being generated before.#3: Inbound Qualification (Higher Complexity, But Massive ROI)SaaStr’s process before Qualified: come to website, fill out contact form, hope over the next week a human gets back to you. Maybe they follow up, maybe they don’t. Round-robin meant inconsistent quality.The Qualified deployment results:* 7 appointments set autonomously in the first weeks* ~100 tickets sold to London event in 6 weeks* Overnight coverage (Pacific team sleeping while UK prospects active)* Better qualification than our human processThe bar wasn’t “be better than a great team of 10 product experts with 5-minute SLAs.” The bar was “be better than broken.” That made deployment possible.What About the “Hero Purchase”?Here’s the failure mode I see CMOs make constantly: they want a hero purchase. They want to go to the CEO and say “I bought AI.”This will probably fail. Why?* That hero purchase is likely aligned with something you’re already decent at* The bar is high because it’s already working* It’s your first agent, so you don’t know how to train it* You’re trying to make something 20% better instead of fixing something that’s 0% doneBetter approach: Leave that hero app on the shelf. Deploy something where work literally isn’t happening. Get a win. Build confidence. Then deploy the hero app when you know what you’re doing.If AI is only going to make something 20% better, maybe 2025 isn’t the year. Maybe that’s a 2026 project.The Daily Reality: What Managing 20 Agents Actually Looks LikeThe hour Amelia spends with agents every morning used to be spent with people.Before: One-on-ones with SDRs and BDRs, going through accounts, providing context on leads, giving them history on prospects.Now: Amelia wakes up, checks each agent:* Qualified — see meetings booked overnight (London prospects while we sleep), review conversations, give context to sales team* Artisan — review outbound performance, adjust sequences* Delphi — check Digital Jason AI conversations, refine responses* Agent Force — review support tickets, catch edge cases* Cycle through remaining agentsTotal time: 1 hourWhat’s different:* Higher cognitive load — you have to think more, use more brain cells* Compounds over time — unlike humans who leave, every hour of training pays forward* Agents don’t cry — they don’t get emotional, but they require deep thinking* You can do more — now expanding to email campaigns, deanonymization, follow-up sequencesThe Early Days RealityFirst 30 days: Way more than 1 hour. You’re setting everything up, fixing data issues (our Salesforce data was terrible — agents expose that), connecting systems, training workflows.After 30 days: Maintenance mode at ~1 hour daily, but you’re constantly expanding use cases.The Compound EffectWith humans: Train for months, they get good, they leave, repeat.With agents: Every hour of training compounds. They get better every day. They don’t forget. They don’t quit. They work 24/7.Example: Qualified handles UK prospects overnight while Pacific team sleeps. Those conversations used to just... not happen. Now they do, and they’re high quality.The Brutal Truths About Team ReactionsFor Founders: Some People Will Leave When You Roll Out AI Agents. That’s OkayAmelia was talking to a founder this week at Dreamforce. She has some agents deployed but is worried about rolling out to her 50-person sales team. “I think they might revolt. Some might quit.”Amelia’s answer: That’s okay. Let them.It’s almost 2026. If your team can’t get over the hurdle, some may quietly leave. You cannot forsake the future of your company to avoid hurting feelings.Rip the band-aid off. Do it the right way — show them they can be empowered with AI, crush quota with AI, do more with AI. The best ones will see this and get excited. They should be roaring to get AI going.If you see people hiding in the corner, maybe it’s okay if they quietly weed themselves out.The Exposure ProblemSaaStr had someone leave their team the day they hooked up Momentum (their conversation recorder and GTM tool). Why? It Slacks summaries after every sales call and notifies Amelia when opportunities open or meetings book.This rep had zero activity. The system exposed it instantly. They quit that day. You can’t hide anymore.For Individual Contributors: Deploy It Yourself or Find a New JobJason recently had a CRO he’s worked with for years reach out in a panicked email: “I need to learn AI. I’ll do anything. I’ll intern for you. I’ll hang out in Amelia’s office.”Jason’s advice: Don’t learn it. DO it.Buying a subscription to Claude does not count as learning AI. Here’s what counts:* Pick one agentic product — simplest possible use case* Buy it yourself — don’t delegate to your team* Set it up yourself — okay to get help with CRM connectors, but own the deployment* Train it yourself — take your time, do something simple* QA and test yourself — hands on keyboardIf you deploy an agent yourself, train it, QA it, test it — you will be ahead of 90% of the world.The CMO Who Jason Couldn’t HelpAnother CMO Jason loves reached out after many years. Jason had gotten her the last two jobs. She was ready for number three.Jason’s response: “I got nothing for you.”Why? She doesn’t know this stuff yet. She hasn’t deployed an agent. She can’t tell Jason how training went, what the issues were, what worked.Here’s the weird thing: There is much more demand for great CMOs than CMOs think. Jason knows 100 CEOs today who are desperate for a great marketer. They’ll do anything.But they do not want someone running the 2019 Marketo playbook. Zero interest in hiring that person.Jason’s advice to her: “Go deploy an agent. Tell me how it worked. Tell me how training went. Tell me the issues. Come back to me. I’ll get you two jobs.”The ROI Reality: What Results Actually Look LikeQualified Results (6-8 Weeks of Deployment)* 7 autonomous meeting bookings per week in early weeks* ~100 tickets sold to London event so far out of ~2,000 total (5% conversion via agent)* 24/7 coverage — overnight conversations with UK prospects while Pacific team sleeps* Better qualification than human process — knows when to escalate, provides context to sales teamArtisan Results (Outbound)* Better email performance than human SDRs who weren’t doing outbound* Consistent execution — no more “round robin to rep who doesn’t follow up”* Scaling sequences we could never scale with humansOverall Impact* From 0 to 20 agents in 10 months* ~12 core agents in production running go-to-market* 1 hour daily maintenance (down from way more in early days)* Work that literally wasn’t getting done now happening 24/7The brutal truth: Half the time, SaaStr’s sales team wasn’t following up with inbound leads. At least half the time. With round-robin, good reps followed up instantly. Mediocre reps... maybe, maybe not.Now: Instant response, 24/7, consistent quality, full context provided to sales team.The Deadline Reality: How Fast Do You Need to Move?For Startups: NowIf you’re a startup and you don’t have anything disruptive in production by now, you’re already behind.Every VC Jason knows, every investor he knows — if the startup hasn’t made the jump yet, they’ve lost hope. Not “losing hope” — they’ve LOST hope.Why?You could make excuses when ChatGPT came out in 2023. There were so many hallucinations. We could all make fun of it.2024? You could be a critic. Maybe you’d start losing deals to hot AI companies, but at least you could have a principled position that this stuff wasn’t going to work.2025? There’s no excuse.Everything’s great now. Even mediocre software got 5x better when you put Claude 4 in. Replit took 10 years to get to $1M ARR. They added Claude 4, now they’re almost $200M in one year. Same IDE. Same software.If you’re not in market with a disruptive AI agent by the end of the quarter:You need a brand new team. Maybe 80% have to go. Give them a nice Thanksgiving bonus and a turkey and tell them they’ve got to go. Because you’ve had 10+ months.For Enterprises: It Depends (But Don’t Wait Long)If your business is fine, if growth is as good or better than 12 months ago, if you’re in a traditional industry where everything’s working — you can be a late adopter.There are late adopter advantages in AI:* Tools get better fast (Replit is epically better than 100 days ago)* Training resources improve* Best practices emerge* Edge cases get solvedBut if you’re facing external pressure, if competition is heating up, if you’re losing deals — you need to move now.The New Software Buying RealityIn the old SaaS days, a customer would come in and say “I need this feature.” You’d promise it in the contract even though it didn’t exist. You’d brute force it over 12 months.Those days are dying in AI.Don’t promise me it’ll be here in a year. Show me the agent works today. If it’s not ready, I don’t blame you. I get it’s early. Come back to me when it works. It might be ready in three months.Time to value has to be before contract signature in many cases now.The Framework: Your 90-Day Deployment PlanDays 1-30: Foundation and First AgentWeek 1: Pick Your Layup* Audit what’s NOT getting done (support? outbound? qualification?)* Pick the simplest use case with lowest bar* Select a leading vendor (doesn’t have to be THE leader, just A leader)Week 2-3: Setup and Integration* Buy the tool yourself (don’t delegate if you’re the exec)* Set up core integrations (CRM, email, support systems)* Clean your data (this will take longer than you think)* Build basic workflowsWeek 4: Daily Training* Train the agent every single day* Test edge cases* Refine responses* Build your QA processSuccess metric: One working agent handling real work by day 30Days 31-60: Expansion and SpecializationWeek 5-6: Monitor and Optimize* Spend 1 hour daily with your agent* Review conversations/outputs* Provide additional context* Expand use cases slightlyWeek 7-8: Agent #2* Pick second use case (slightly harder)* Apply lessons from Agent #1* Faster setup because you know what you’re doing* Stair-step to more complexitySuccess metric: Two agents in production, 1 hour daily maintenanceDays 61-90: Scale and SpecializeWeek 9-10: Vertical Agents* Move from horizontal to specialized* Deploy agents for specific workflows* Start tackling harder use cases* Build your agent management systemWeek 11-12: Team Enablement* Train your team on working with agents* Build handoff processes* Create monitoring systems* Document what worksSuccess metric: 3-5 agents in production, team enabled, clear ROIThe Future: What’s Coming NextVoice and Video Agents (Live at SaaStr Now)SaaStr is deploying Amelia AI with voice and video. Not just text chat — full video agent of Amelia on their event sites.Why? Trust has been built through chat. Now the team can do more:* Overnight support with full video presence* Event co-pilot for attendees* Session recommendations* Networking coordination* “Real Amelia is asleep, but Amelia AI is awake”The funky part: People walk up to real Amelia now and say “You’re that Amelia AI!” The line between digital and real is blurring.The Compounding Use CasesOnce you have 5-10 agents working well, new use cases emerge constantly:* Email deanonymization of website traffic* Follow-up campaigns for unconverted prospects* Pre-meeting research and briefings* Post-meeting follow-up and deal progression* Internal knowledge management* Onboarding automationThe cognitive load increases, but so does output. It’s exhausting in a different way than managing people, but it’s more productive.The Market Reality: Everyone’s In MarketTraditionally, you’d hope a prospect would be in market every 5-6 years for a big buy. Why? Takes a year to get going, second year to scale, third year to get frustrated, fourth/fifth year to start looking at new solutions.Right now, everything’s in market again.This hasn’t happened since maybe when the web started. Everyone either has AI budget or is under pressure from their board/CEO to show AI innovation. It’s the only incremental budget.For vendors: This is the opportunity of a lifetimeFor buyers: The tools will never be better-priced or more hungry for your businessThe Final Word: Just Rip the Band-Aid OffLet’s be brutally direct:If you’re a go-to-market executive and you haven’t deployed an AI agent yourself by now, you’re unemployable.Not “will be unemployable” — you ARE unemployable. Jason has CMO friends he loves, who he’s gotten jobs for multiple times, who he now has nothing for. Why? They don’t know this stuff yet.The Simple Advice:* Don’t learn AI — DO AI* Pick one agent, simplest possible use case* Deploy it yourself (hands on keyboard)* Train it for 30 days* Spend 1 hour daily after thatYou will be ahead of 90% of the world.The agents don’t cry. They work 24/7. Every hour of training compounds. They get better every day, they don’t forget, they don’t quit.But only if you actually do it.It’s almost 2026. Rip the band-aid off. Let’s go.Key Takeaways* Training > Vendor Selection — Pick a leader and go deep, training matters more than choosing the perfect tool* Stair-Step Your Deployment — Start simple, build confidence, then tackle harder use cases* Find the Layup Roles — Deploy where work isn’t getting done, not where it’s already working* Budget 30 Days + 1 Hour Daily — This is your new reality, embrace it* Halloween Deadline for Startups — If you’re not in production now, you need a new team* Do It, Don’t Learn It — Hands on keyboard deployment is the only way to actually learn* Some People Will Leave — That’s okay, you can’t forsake your company’s future* The Compound Effect is Real — Unlike humans, agents get better every day foreverThe future belongs to companies that move now. Not next quarter. Now.Want to see Jason AI in action? Go to SaaStr.com and click the bottom right corner. Want to see how SaaStr deployed 20 agents? Follow Jason Lemkin on LinkedIn or come with your questions LIVE to SaaStr AI London on Dec 1-2!! This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

I think after 100+ days and with 9 apps vibe coded into production with @replit used over 500,000 times we’re just getting going. And … I think key to that is that I’m now “Replit Fluent”.What does that mean? It’s a state where I know how to vibe well enough (without a developer), and I know the app and its capability and limits well enough, that I can basically see any app I want to build in my head, and now know how to completely prompt it and shepherd it to production … before I start.I can now will almost any ‘normal’ app into existence that I want to build. Without a developer.I remember in the early days of Cursor my son actually paid out of his own pocket for the first time since ChatGPT (he’s awfully smart). He said back then Cursor could now do 90% of his coding for him, but “for most people, it might be 10% or less.” I didn’t get what he meant at the time, but today I do. It’s more than being “good at prompting”. It’s understanding the system well enough to already understand its outputs, its limitations, and exactly how it works in practice — before you start. It’s being truly fluent in the agent.This isn’t to say I don’t still have bumps and that some things end up harder or longer than anticipated. But pretty much now I can will most things into production on Replit very predictably with the agent — because I can already plan them out fully in my mind before I first “prompt” the agent.What that means in practice is 3 things:First, I 100% know if a project will work now before I start in Replit. If I can see it to completion in my mind, now I can finish it to a reasonably high standard. That is a huge boon. When I started ‘prosumer’ vibe coding all of 100+ days ago :), I couldn’t finish my first project. In part, it was because I picked a very complex product to start. But there are many stories of others in similar boats. They can’t finish their vibe coded app. But now — I have a 100% chance of finishing a project, and in roughly the amount of time I budget for it in mind before starting.Second, now I’m merely just time constrained in what I can build. I already have a couple of jobs. I have $100m+ to invest in fresh capital at SaaStr Fund, and running SaaStr itself is an eight figure business. Both take a lot of time. But I set aside about 1.5-2 hours a day to vibe code. That’s my budget. That’s what I can build now.Third, maintenance and new features and upgrades to existing apps I’ve vibe’d consume more and more of my time budget. This of course is true of any software. It just catches up to you with prosumer vibe coding. So now that I can basically truly build anything I want, the question becomes — do I have enough time to make it great? Or should that time go into making my existing 9 apps even better? I now can make a pretty good v1 of anything I want to. But getting to great takes time. It always has.I think it’s a big deal to be able to get to this state without a developer and without being a developer. Yes, I co-founded a B2B startup that went from $0 to $200m+ ARR so I had that going for me (which is a lot). But the fact I am now “Replit Fluent” and that now my only limit on building is my own time … is super interesting.The world is so, so much different than 12 months ago.There is a learning curve in AI. In almost every great tool and every great agent, at least for now. Don’t let anyone tell you there isn’t. Ignore any marketing that tells you there isn’t. If you want to get most AI Agents into production, e.g. a great AI SDR or great AI Support, and have it work well — you have to learn and often train the AI. IMHO you also have to learn tools like Replit to truly get out of them what they promise. You will be much, much better 100 hours in.But if you get to a State of Fluency in a prosumer vibe coding app … it’s like a superpower you never had before.Thanks for reading SaaStr AI: How To Sell, Scale, and Win! Subscribe for free to receive new posts and support my work. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

Marchelle Mooney, VP of Sales at Mangomint joined us at 2025 SaaStr Annual + AI Summit for one of our best deep dives on sales and GTM in vertical SaaS yet.Marchelle brings a unique perspective to vertical SaaS sales—she’s a former hairdresser and salon owner who transitioned into SaaS sales leadership. At Mangomint, she leads sales for a deep vertical SaaS platform serving salons and spas. Her journey from thinking “SaaS just meant you had attitude” to becoming a VP of Sales at a fast-growing vertical SaaS company gives her insights that bridge the gap between traditional enterprise SaaS playbooks and the reality of selling to SMB vertical markets. She credits much of her learning to Jason Lemkin and the SaaStr community.MangoMint today processes over 1,000,000 appointments a month for over 5,000 salon and spa customers.Top 5 Learnings: The Vertical SaaS Sales Playbook1. Deep Domain Credibility is Non-Negotiable—Build Trust From the First ClickThe insight: Your founding team needs deep insider knowledge of the vertical before shipping a single line of code. This isn’t about market research—it’s about speaking the language fluently.Marchelle emphasizes that credibility builds trust instantaneously. From the moment a prospect lands on your website, they should see language that feels like their own world. At Mangomint, this means understanding the difference between how a hair salon refers to their “client” versus how a med spa calls them a “patient.”She’s witnessed demos end immediately because a rep used the wrong terminology. In one case, calling a “patient” a “client” was enough for the prospect to disconnect—that’s how critical getting the language right is in vertical SaaS.The key principle: You’re solving problems for customers who don’t know what they don’t know. Many verticals have been using antiquated playbooks for years. Your product needs to answer questions they haven’t even thought to ask yet, and you can only do that with genuine insider knowledge.Why it matters: When Marchelle started at Mangomint in 2018, every tenth call involved unseating pen and paper booking systems. Today, it’s maybe once a year. The market has evolved, but the need for domain expertise hasn’t—it’s just shifted to different problems, like introducing AI to customers who’ve only used ChatGPT to decide between sushi or Thai for dinner.2. Win by Eliminating Choice—Your SMB Customers Have Analysis ParalysisThe insight: Vertical SaaS customers, especially SMBs, struggle with too many options. Unlike enterprise buyers who might want flexibility and customization, your vertical customers want you to tell them the right way to do things.Marchelle uses a simple analogy: “What are we doing for dinner? We could do sushi. We could do Thai.” As soon as there are two options, she’s in analysis paralysis. Her customers feel the same way.The approach: At Mangomint, they’ve eliminated choice not by limiting features, but by presenting a clear, opinionated path forward. They’re going after a vertical where customers aren’t using any AI products except ChatGPT for personal decisions. This creates an opportunity to be prescriptive about the right solution.Why this works in vertical SaaS: Your customers are busy operators. A salon owner is literally doing a demo while bleach is processing on a client five feet away, and they might need to pause the call to shampoo someone. They don’t have time to evaluate ten different workflow configurations—they need you to show them the one that works.3. PLG + SLG = The Perfect Vertical SaaS Motion (When Done Right)The insight: The debate about “PLG is dead” versus “SLG is back” creates a false dichotomy. In vertical SaaS, you need both—but PLG serves a different purpose than in traditional SaaS.Mangomint offers a 21-day free trial (not a freemium product), but they don’t use it to replace sales—they use it to supercharge sales conversations.How it works:* Prospects can start a trial and explore the product independently* But the trial becomes a data goldmine for sales reps* When Sally from the waxing studio goes straight to the memberships tab, rage clicks, and leaves, the AE gets an alert* The rep sends a personalized text: “Hey Sally, I’m Marshelle with Mangomint. Thanks for starting a trial. Looks like you’re building memberships. What memberships do you currently offer? Would you like to see how Jessica’s spa down the street does $30k a month in memberships using Mangomint?”The framework: “The product proves, the humans frame the value.”The trial isn’t about product-led conversion—it’s about conversation-led conversion informed by product data. This is personalization at scale, using PLG as a tool rather than a replacement for sales.4. Master Sub-Vertical Dominance Through Micro-Niche FluencyThe insight: Within your vertical, there are sub-verticals, and each one thinks they’re unique and special. Treat them like the unicorns they believe they are.At Mangomint, a hair salon owner doesn’t use the same terminology as a med spa owner. These aren’t just semantic differences—they represent different mental models, different workflows, and different Facebook groups (literally—hair salons don’t join the same Facebook groups as med spas).The strategic approach: Marchelle calls this “breaking through the silence.” Unlike horizontal SaaS where you’re cutting through noise, in vertical SaaS you’re often dealing with silence—isolated communities that don’t cross-pollinate. Each sub-vertical lives in its own bubble.The execution: Go on “sub-vertical dominant sprints.” Master one micro-niche completely before expanding. Know the words they use, train your reps with 100% fluency, and dominate that sub-vertical before moving to the next.Why this matters: This isn’t about having a tighter ICP or a cuter logo—it’s about fundamentally different go-to-market execution. The playbook is flipped upside down from traditional SaaS.5. Make Onboarding Your Sales Weapon—Trust Before RevenueThe insight: In 2022, Mangomint moved their customer success team into the new revenue organization and transformed them into an onboarding team. This wasn’t just an org chart change—it was a strategic repositioning of how they build trust.The approach: Mangomint offers completely complimentary white-glove onboarding. They migrate client lists, client history (sometimes medical history), sales history, service menus, staff schedules, and payroll setup—all for free, often before the customer has paid a single dollar.Why SMB vertical SaaS is different: These customers don’t want to pay to migrate their mission-critical data. Unlike enterprise customers who might have budgets for implementation services, SMB customers in vertical markets expect you to make it easy.The payoff: By building trust first through free onboarding, you can later layer in charging for premium outcomes and growth features. Trust is Mangomint’s number one company value “by design, not by default.”This approach also extends to support. While many SaaS companies are moving to AI-first support, Marchelle heard the HubSpot CEO and others say “support is done, it’s AI-owned” multiple times at SaaStr. But when she asked her VP of Customer Operations about their AI support tool Finn, the response was revealing: “Finn doesn’t know who canceled the haircut on Tuesday at 2pm with Jessica.”The vertical SaaS reality: You need a different support playbook. Problems that come up might not be software problems—they’re vertical problems. You need to show customers you can help them with these issues and teach them to solve them over time. (Good news: They eventually have a report for that, so the person won’t reach out again about canceled appointments.)Top 3-4 Mistakes Marshelle Made (and What She Learned)1. Initially Underestimating the Importance of Vertical Athletes in HiringThe mistake: Early on, there was likely too much emphasis on finding reps with SaaS experience and big logos on their resumes, without enough weight on their ability to operate at the altitude of the vertical.The learning: A Salesforce enterprise rep with an impressive resume might be completely wrong for selling to a salon owner doing a demo while processing bleach. The mismatch isn’t about capability—it’s about altitude and agility.The correction: Marchelle now focuses on hiring “vertical athletes”—reps who are agile enough to understand the customer and operate at their level. While she still values SaaS sales acumen, the persona is most important: “I’m looking for the Grade-A top tier USDA approved beef, grass-fed, all of it.”She notes that all their top performers now have prior vertical experience, but she’s confident they could perform in other companies too. The key is being holistic about team building, not fixating on either domain expertise OR SaaS experience—you need both.Current approach: She’s now focused on immersing reps in the trade during onboarding and exploring AI simulators that can plug in actual conversations from the vertical, allowing reps to practice until they’re fluent in the language their customers speak.2. Not Moving Fast Enough on Trial Response TimesThe mistake: Not initially understanding that the window for connecting with business owner/operators is measured in minutes, not hours or days.The learning: When a salon owner is taking 30 seconds out of their busy day to search for software and start a trial, they’re looking RIGHT NOW. If you don’t call them within minutes of trial signup, you’ve probably missed the initial window.The correction: Marchelle implemented a strict SLA: contact within minutes, with no exceptions. The sequence is automated message + phone call + text follow-up—always from a human, always immediately.Why this matters differently in vertical SaaS: Unlike traditional SaaS buyers who might be doing structured vendor evaluation, vertical SMB customers are often impulse shopping during a brief break in their operational chaos. Speed isn’t just important—it’s everything.3. Trying to Be Everything to Everyone (Learning to Say No to Non-Ideal Customers)The mistake: Getting caught up in the hype cycle and not proactively disqualifying prospects who aren’t a great fit, particularly franchises in Mangomint’s case.The learning: Growth loves focus, and this is even more critical when you’re trying to be the default leader for a specific vertical. Being “proudly not for everyone” isn’t just positioning—it’s a crucial strategic decision.The correction: Now, if Marchelle sees any indication that a prospect might be a franchise (something that “gives her an itch on her spine”), she proactively has that conversation and lets them know Mangomint isn’t great for them—before they self-identify this during the trial.The framework: Don’t push into markets that aren’t pulling you there. Stay niched down enough that you can be genuinely best-in-class for your ICP rather than mediocre for a broader market.4. Misunderstanding What “Customer Success” Means in Deep Vertical SaaSThe mistake: Initially thinking about CS as a separate team function rather than as a company-wide responsibility.The learning: In vertical SaaS, you can’t afford to have CS as a siloed function. The problems customers face are too intertwined with product, sales, and operations.The correction: Mangomint “killed their CS team” but made customer success more important as an organizational principle. They implemented shared ownership across the entire company:* Product owns success metrics and time-to-value* Sales owns reality—helping customers understand both what the product can and can’t do for their specific situation* Customer Operations (live in-app support) owns the inevitable problems that ariseThe vertical SaaS insight: This isn’t just a “nice to have” or aspirational—it’s foundational. When you’re serving very specific verticals with deep operational needs, everyone needs to be aligned on customer success, because the problems don’t fit neatly into traditional support categories.The Big TakeawayMarchelle’s final point encapsulates the entire vertical SaaS philosophy: Don’t be just another software—be the only software they’ll need.Mangomint’s mission is to make every salon and spa more profitable. Right now, they’re focused on being best-in-class, the default software for their ICP. But down the line, they’ll be the software that teaches every salon and spa to bolt on AI.In the future, customers will likely book appointments through AI agents—but Mangomint will get there by building trust first, ensuring their vertical feels safe migrating in that direction, and making sure customers don’t feel like they’re ripping out the human element of their business.As Macshelle jokes with a woman working the speaker room about having “a job AI cannot take,” only to learn about a shop down the street where robots are doing eyelashes—the future is coming. But in vertical SaaS, you guide your customers there. You don’t just sell them software. You become their trusted partner for navigating change in their industry.That’s the vertical SaaS playbook. It’s not just a tighter ICP or a cuter logo. It’s fundamentally different from the ground up.Thanks for reading The Secrets To Scaling in The Age of AI! Subscribe for free to receive new posts and support my work. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

The VC fundraising landscape has completely transformed in the last 18 months, and most founders still don’t realize just how dramatically the rules have changed.After analyzing 1,000 VC pitch decks and calculating 400,000+ startup valuations on SaaStr.ai, and having countless conversations with both sides of the table, the data is unambiguous: traditional paths to venture funding have essentially closed for the majority of B2B companies. The capital isn’t gone—there’s actually more money in the market than ever. It’s just almost all flowing to a completely different type of company than it was 24 months ago.Jason Lemkin joined us for a LIVE SaaStr AI Wednesday to walk us through the data, and do live Q+AIf you’re a B2B founder growing 80% annually at $10-30M ARR with solid unit economics and happy customers, you might think you’re in a strong position to raise. You’re probably not. If you’re planning to raise your Series B based on performance that would have secured funding in 2022, you need to recalibrate immediately. Benchmarks have shifted so dramatically that roughly 80% of VCs who would have enthusiastically funded solid B2B companies 18-24 months ago are now passing—not because those companies got worse, but because an entirely new category of hypergrowth AI-native startups has reset every expectation in the market.This deep dive breaks down exactly what’s happening, why it’s happening, and—most importantly—what you need to do about it right now. We’ll walk through the actual data from top-tier growth funds, show you the real benchmarks you’re being measured against, and give you a clear-eyed assessment of your options whether you’re pre-revenue, scaling past $10M, or approaching $100M ARR.The worst thing you can do is stay in the dark about where you actually stand. Let’s fix that.Top 5 Takeaways* The bar for VC funding has skyrocketed dramatically. AI-native companies are now scaling from $1M to $100M ARR in 8-11 quarters (versus the previous “top quartile” benchmark of 19-20 quarters), fundamentally resetting investor expectations across the board.* 80% of traditional B2B VCs who would have funded you 18-24 months ago won’t fund you today. Capital is flooding into hypergrowth AI-native companies, leaving traditional SaaS startups—even strong ones—struggling to raise regardless of solid fundamentals.* The latest top quartile metrics are higher. And harder. At $1m-$5m ARR, VCs expect 500% growth. At $10-25M ARR, VCs expect 100%+ growth. At $50-100M ARR, they want 90% growth with 120% net revenue retention. Below these numbers, you’ll face an uphill battle regardless of market or team quality.* Three consecutive great months can flip you from unfundable to fundable. If you’re close to the benchmarks but not quite there, demonstrating acceleration over 3-4 months can completely change investor perception—but you need to be honest about where you actually stand first.* Don’t assume the next round is coming. The single biggest founder mistake is burning cash with the expectation they’ll raise another round. If your odds are below 60-70%, run your company with the cash and revenue you have—not the funding you hope to get.The New Reality: AI-Native B2B Speed vs. Traditional SaaS GrowthLet’s start with the chart that explains everything happening in VC right now. Iconiq Capital, one of the leading growth funds originally backed by Mark Zuckerberg’s money, recently published benchmark data that crystallizes the entire market shift.The Old Top Quartile: Getting from $1M to $100M ARR in 19-20 quarters (roughly 5 years) used to represent elite performance—the kind of company VCs actively competed to fund.The New Standard: AI-native companies are doing it in half that time or less:* Glean (AI enterprise search): 11 quarters* 11 Labs (AI speech): 8 quarters* Perplexity: Even faster* Cursor, Anthropic, Lovable, Replit: Less than 4 quarters from $1M to $100MHere’s the question every VC is now asking: If you have $100M, $500M, or $2B to deploy, where on this spectrum are you putting your money? The answer is obvious for most—and it’s not with the companies on the right side of that chart, even though they’re very strong “top quartile” performers.These hypergrowth AI companies didn’t even exist 24 months ago. You literally couldn’t invest in them. So the top quartile traditional SaaS companies were about as good as it got. But now? That entire world has been reset.Two Categories of Fundable CompaniesBessemer Venture Partners broke down their investment thesis into two distinct categories, and the contrast is striking:Supernovas (AI-Native Companies)* Growth expectations: $1M to $40M in year one, $125M+ by year two* Gross margins: Often 25% or even negative (yes, negative)* Revenue per employee: $1M+* Examples: Gamma hit $50M ARR in one year; Replit and Lovable are burning massive cash but scaling at unprecedented ratesShooting Stars (The Best of Traditional Software Startups)* Growth expectations: $1M → $3M → $12M → $40M → $100M (still better than triple-triple-double-double over 5 years)* Gross margins: 60%+ required* Revenue per employee: $164K minimum* Key point: Even for these “traditional” companies, the bar has never been higherThese AI-native companies often have terrible gross margins because of token costs and infrastructure expenses. But VCs don’t care. They’re remarkably headcount-efficient (often under 100 employees at $50-100M ARR), and the growth trajectory is so explosive that margin concerns become secondary.The Paradox of Public vs. Private ExpectationsHere’s where things get confusing for founders: while VC expectations for startups have skyrocketed, expectations for public B2B companies have actually declined.Public company reality:* Average growth rates are at all-time lows* “High growth” for public companies is now just 30% (it was 70%+ in 2021)* Companies growing 20-30% are getting 20x+ ARR multiples* HubSpot is growing 19% and trading at premium valuationsPrivate company reality:* VCs expect 150-250%+ growth depending on stage* The bar keeps rising, not falling* You need to massively outperform public company benchmarks to even get a meetingWhy the disconnect? Public markets value profitability and predictability at scale. Private markets are chasing the next category-defining winner, and AI has created more potential winners than ever before—meaning the competition for capital is fiercer despite there being more capital available.The Real Benchmarks: What Top Quartile Actually Looks Like NowFrom Iconiq’s data analyzing their portfolio of venture-backed B2B companies, here are the top quartile benchmarks you need to beat to get funded:Less than $10M ARR:* Growth: 515%* Net Revenue Retention: N/A (too early)* Magic Number: N/A* Revenue per employee: N/A$10M-25M ARR:* Growth: 170%* Net Revenue Retention: 130%* Payback period: 1.0x* Revenue per employee: $160K$25M-50M ARR:* Growth: 100%+* Net Revenue Retention: 125%* Payback period: 1.2x* Revenue per employee: $180K$50M-100M ARR:* Growth: 90%* Net Revenue Retention: 120%* Payback period: 1.5x* Revenue per employee: $220KRead that carefully. This is not “top 1%” or “exceptional” performance. This is top quartile—meaning 75% of top venture-backed B2B startups fall below these numbers. And yet, this is effectively the minimum bar to get funded by most VCs today.If you’re at $30M ARR growing 60-70%, you are building a truly great company. You might even have excellent fundamentals, happy customers, and strong unit economics. But you’re walking through the desert when it comes to VC funding. Maybe—maybe—you can raise if you have extraordinary circumstances (dominant market share, insane NRR, clear acceleration path), but you’ll need exceptional reasons beyond being likable and having a solid business.The Data from 1,000 Pitch Decks and 400,000 ValuationsThrough SaaStr AI’s pitch deck review tool (which has now analyzed 1,000+ VC pitch decks in just two weeks), we’re seeing the market reality play out in real-time:Series A averages (across ~108,000 calculations):* AI-native companies: ~$5-6M ARR, growing 180%* AI-enhanced SaaS: ~$5M ARR, growing 100%* Traditional SaaS: ~$5M ARR, growing 75%The traditional SaaS number (75% growth) won’t get you funded in most cases. It’s just not enough.Grade distribution from pitch deck reviews:* 17% received A- or above (these are “top tier” and highly fundable)* Average grade: B-* Median grade: B-Here’s what that means: Most founders are in the B range—they have something, they’re in the zone, but it’s not enough. A B- is not enough to get funded in today’s market. You need to be brutally honest about where you actually stand.What Revenue Efficiency Really Looks LikeOne trend that’s accelerating: companies are getting dramatically more headcount-efficient as they scale. The Iconiq data shows this progression:* Early stage: ~$90K revenue per employee* Growth stage: $220K revenue per employee* Pre-IPO: $400K revenue per employeeThe most successful companies—especially AI-native ones—are proving you can get to $50-100M ARR with under 100 employees total. That level of efficiency is becoming the new expectation.How to Navigate This Market: Practical Advice#1. Get Brutally Honest Feedback on Your Actual Funding OddsDon’t go into fundraising blind or optimistic. Use these approaches:Ask your existing investors directly: Put them on the spot. “On a scale of 1-10, how good is our startup? More importantly, what are the exact odds—give me a percentage—that we can raise our next round?” Push for a number. Most experienced investors know exactly what your odds are.Use objective AI tools: Upload your pitch deck to SaaStr AI’s review tool. It will benchmark you against all other founders, tell you where you stand, show you what VCs will think of your traction, and give you honest odds of getting funded. Don’t let ego or optimism cloud your judgment—you need to know where you actually stand.#2. Figure Out What You’re Number One AtThis has always mattered in venture, but VCs are obsessed with it now. With hundreds of AI SDR tools, thousands of AI startups, and constant innovation noise, you must be number one at something specific.Not “we’re going to be better than the other guy eventually.” Not “we have a great team and we’re executing well.” What are you number one at right now? And show it with data, even if you’re early.Every VC is looking for the next category winner, the next OpenAI of their vertical. If you can’t articulate—and prove—what you’re number one at, you’ll get lost in the noise.#3. Understand Your ZoneIf you’re above the Iconiq benchmarks: You’ll likely get funded. Maybe not from everyone, but you’ll get multiple term sheets if you have good founders and an interesting space.If you’re within 20% of the benchmarks and burning very little: You’re in the “maybe” zone. Not everyone will want to invest, but you’ll find someone—especially if you have other compelling factors (new product coming, obvious acceleration path, strong market position). Just be prepared to take a lot of meetings.If you’re significantly below the benchmarks: Be honest. It’s going to be incredibly difficult if not impossible. Don’t pretend otherwise. Don’t run out of money because you assumed you could raise. The worst outcome in startup life is going from funded to unfunded and not knowing it until you’re out of cash.#4. The Three-Month RuleIf you’re close but not quite fundable, here’s the good news: VCs have incredibly short memories for previous underperformance if you can show recent acceleration.If you’re at $20M ARR growing 70% (below the 90% benchmark), but people are taking meetings because they believe in you and your space, three consecutive great months can completely flip the narrative.Show them one month where instead of growing 6% MoM you grew 10%. Then another month at 10%. Then a third at 10%. By month three or four, investors will stop caring about the previous 30+ months of slower growth.You don’t need a year of great performance. You don’t need 10 years. But you probably do need three solid months, maybe four to be safe. Two months of acceleration will get you meetings, but investors will want to see one more month before committing.Define a clear inflection point (launched AI agents, hired key exec, released version 3.0), explain what changed, and show sustained acceleration from that point forward.#5. Not Everyone is AI-Native or Bust (But 80% Are)The VC world has split into rough camps:~80% of VCs: Only interested in AI-native, hypergrowth companies. If you’re not building the next Replit or Cursor, they’re out.~20% of VCs: Still believe in traditional triple-triple-double-double SaaS. These investors (like Rory on the 20VC podcast) argue that if you can hit those benchmarks, they can see a path to IPO at multi-billion dollar valuations, and that’s good enough to make money.The takeaway: Don’t jump off a bridge if you’re not AI-native. You just need to realize the fundraising dating game will be harder. You’ll need more meetings, more outreach, and thicker skin. But the investors exist—they’re just the minority now.#6. Don’t Assume the Next Round is ComingThis is perhaps the most critical point: The biggest mistake founders make is assuming they’ll be able to raise another round.Look, if you’re at $20M ARR growing 60%, you’re still compounding toward something potentially great. If you don’t need outside capital and can grow at 50-60% for several years, you’ll build a hundreds-of-millions-dollar company. That’s objectively a great outcome.But no VC will fund you at those metrics in today’s market.So stop operating like the next round is guaranteed. Stop burning cash with the assumption you’ll raise. If your odds are below 60-70% (and be honest—they probably are if you’re not hitting those benchmarks), run your company on the cash and revenue you have.This isn’t bootstrapping shame. This isn’t giving up. This is basic financial reality. Either you can raise money to run at a loss in exchange for hypergrowth, or you can’t. If you can’t, don’t run at a loss.#7. Be Wary of Dated AdviceThis is subtle but important: many investors haven’t been actively investing recently, or they’re disconnected from what’s actually getting funded in 2025. They’ll give you advice that sounds reasonable but is actually from 2021 or 2022.They’ll tell you your company is “hot as Travis Kelce and Taylor Swift” because you have great unit economics and a strong team. But if you don’t have an ounce of AI in your product and you’re growing like a 2021 company, nobody will actually fund you at the valuations you expect.The same goes for executives you hire. Be extremely careful about VPs of Sales or CMOs who haven’t been in the market since 2021. They’ll want to build teams of 20-40 people, hire 12 people in finance, create 21-person demand gen teams. That model is increasingly dated and unsupportable.Get advice from people who are actively in the market right now. Not people who were successful three years ago but haven’t adapted to the current reality.Special Considerations for Different StagesPre-Revenue and Pre-Seed CompaniesIf you’re pre-revenue or pre-seed, the signaling game becomes even more important:Join a top accelerator if possible: Y Combinator, Neo, South Park Commons, or Entrepreneur First. Not because the programs themselves will make you a better company (though they might), but because the investor attention is at an all-time high. Yes, it’s somewhat unfair that 20-year-old founders with two weeks of revenue and massive hubris can close rounds at $25-60M post-money valuations just because they’re YC-backed. But that’s the reality. These accelerators attract VCs like flies.If you can’t do an accelerator, you need alternative signals:* Epic team (proven founders or engineers with strong reputations can raise pre-revenue)* Exceptional pilots (three pilots that could each become multi-million dollar contracts)* Early traction metrics (100K users for a free product, 10 real paying customers, something demonstrable)The “Almost There” CompaniesThis might be the toughest spot to be in right now: You’re at $10-20M ARR, growing 50-70%, maybe even cash flow positive or close to it. Through 2023, this is exactly when private equity firms would start calling to buy you. Companies at $15-30M ARR growing 30-50% were getting acquired at 6-10x revenue multiples.That’s largely stopped. PE firms learned that many of these acquisitions (made in 2020-22) haven’t played out as expected. Revenue hasn’t proven as durable, AI is making categories unstable, and growth has decelerated faster than models predicted.What to do:* Take PE calls if they come, but don’t expect them like you would have 18 months ago* Push forward and get bigger—the PE desert between $10-30M ARR is real right now, but it may not be permanent* AI-enhance your product aggressively so you can potentially access either VC funding or PE acquisition in 12-24 months with better metrics* Study companies like Filevine (legal case management) that infused true AI agents and raised $400M from Excel at a $3B valuationThe Professional Services QuestionFounders often worry about having 50% professional services revenue versus 50% software ARR. Don’t.Palantir—worth over $300 billion today—was almost entirely low margin software + services until just before going public. Their gross margins were in the teens or 20% range … until suddenly they weren’t. Every VC will tell you services are bad, but Palantir literally proved the opposite at massive scale.More importantly, “forward deployed engineers” are the hottest hiring trend right now (up 12x in the last year according to Iconiq). Why? Because AI products require training, configuration, and month-long deployments. This isn’t technically professional services, but it’s close.Don’t worry about the revenue mix. Worry about whether you have real product-market fit and can scale very rapidly.The SAFE vs. Equity QuestionQuick primer: Most VCs don’t want to do SAFEs unless the check is immaterial or they have to.SAFEs offer investors no protection, no control over dilution, limited (if any) pro-rata rights, and no board seat. There have been cases where founders simply kept all the SAFE money because legally “it’s neither equity nor debt, so investors have no rights.”When do SAFEs make sense?* Large funds writing tiny checks ($100K from a multi-billion dollar fund—they don’t care)* Forced by accelerators (YC makes you use SAFEs; everyone accepts it for small amounts)For a $2M raise, you might be able to do it on a SAFE, but unless you have a hot hand, don’t push it. Optimize for getting the deal done, not the specific instrument. Don’t over-optimize on terms that don’t really matter if it means losing the deal.Understanding Valuation ExpectationsHere’s an uncomfortable truth: Not only have VC expectations gone up, but founder valuation expectations have also skyrocketed—especially for AI companies.Recent Y Combinator batches have had founders with minimal revenue raising at $25-60M post-money valuations, hoping to triple that in another round within a year. From the VC side, that’s an extraordinarily difficult bet to make work economically.So you have VCs expecting faster growth and founders expecting higher prices with lower dilution. This puts pressure on the entire model. You might get the term sheet, but the performance expectations will be equally extreme.The Bottom LineThe venture capital market in 2025 has fundamentally bifurcated:The winners: AI-native companies scaling at unprecedented rates (even with negative gross margins), and exceptional traditional SaaS companies hitting 100%+ growth at $25M+ ARR with strong margins and capital efficiency.Everyone else: Walking through a funding desert, regardless of having solid businesses, happy customers, or reasonable growth rates.This isn’t fair. It isn’t rational in many ways. A company growing from $20M to $35M in a year (75% growth) is still building something substantial and valuable. But VCs are human, capital is following the most explosive growth stories, and you can’t fight the market.What you can do is be ruthlessly honest about where you stand, understand your actual odds, make informed decisions about burn rate and runway, and either position yourself to hit the benchmarks or run a sustainable business without assuming the next round will materialize.The worst outcome isn’t building a $100M revenue company over 10 years instead of 3 years. The worst outcome is running out of money because you didn’t realize the rules had changed.Top Things Founders Need to Know Right Now* Use objective tools to assess your fundability. Don’t rely on gut feel, friendly advisors, or outdated advice. Upload your deck to SaaStr AI, ask your investors point-blank for percentage odds, and benchmark yourself against current market data—not 2021 metrics.* If you’re not AI-native and not hitting top quartile growth, prepare for 5x more meetings. You’ll need to talk to everyone. Get comfortable with rejection. Have a thick skin. The investors who will fund you exist, but they’re now the minority.* Three great months can change everything. If you’re close but not quite there, focus obsessively on showing acceleration over 90-120 days. Define your inflection point clearly, show consistent execution, and investors will forgive previous underperformance.* Make the cash you have last. Run your company as if the next round isn’t coming unless you’re highly confident (>70% odds) you can raise. The biggest failure mode is burning through capital while assuming funding is inevitable when it’s not.* Be number one at something specific and provable. “We’re going to be great” doesn’t cut it. “We have 10x the accuracy of competitors as measured by [specific metric]” or “We’re the fastest-growing solution in [narrow category] with [specific data]” is what VCs want to hear.* Don’t confuse PE recapitalizations with venture rounds. If someone wants to give you $20M but take 50% of your company, that’s not growth capital—it’s a partial buyout. These deals exist and might make sense, but they’re fundamentally different from venture funding.* The bar is high and getting higher, but it’s not arbitrary. VCs are comparing you against the fastest-growing companies in history. They’re not trying to be mean—they’re trying to make returns in a market where AI-native companies are redefining what’s possible. Your job is to either meet that bar or build a great business without venture capital.The market has spoken. The only question is whether you’re going to listen and adapt accordingly.For more resources, benchmark your company for free at SaaStr AI (saastr.ai), where you can get instant valuations, pitch deck reviews, and see how you stack up against thousands of other B2B startups in real-time.Thanks for reading The Secrets To Scaling in The Age of AI! Subscribe for free to receive new posts and support my work. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

We’re back! Our latest deep dive with Jason Lemkin (SaaStr), Rory O’Driscoll (Scale Venture Partners), and Harry Stebbings (20VC).And come join the team LIVE at SaaStr AI London, Dec 1-2!! More here.Bottom Line Up FrontThe classic venture playbook is almost … no more.* Traditional SaaS metrics like burn multiples—once the gold standard for evaluating capital efficiency—are being rendered obsolete by AI-native companies growing at unprecedented speeds with radically different unit economics.* Meanwhile, founders with objectively good numbers (triple-triple-double-double growth, solid burn multiples) are getting rejected by VCs focused exclusively on AI breakouts.* The message is stark: if you’re not AI-first, raise capital now at any reasonable price, consolidate where possible, and prepare for a world where even “perfectly good” $15M ARR companies have “zero value to VCs.”The stakes extend beyond individual companies:* With 6,700+ unicorns and only 15 IPOs year-to-date, massive portfolio consolidation is inevitable.* Tech PE firms face existential questions as AI agents reduce seat-based revenue and products that stayed static for a decade now require constant reinvention.* And towering above it all: OpenAI’s plan to consume more energy than India within 8 years, raising fundamental questions about whether the economics of AI can ever pencil out at scale.The Burn Multiple Paradox: Why AI Breaks the RulesThe conversation opened with Iconiq’s 73-page State of Software report, which revealed a counterintuitive finding: AI-native companies under $100M ARR have terrible free cash flow margins (-126% versus -56% for non-AI companies), yet their burn multiples are actually better because they’re growing so explosively fast.Rory O’Driscoll explained the fundamental concept: “It’s basically how many dollars of ARR do you get out of each dollar that you’re spending, right? What is the efficiency you’re creating for each dollar of venture capital you’re lighting on fire?”The math seems simple: if you’re valued at 10x ARR and you spend $2 to add $1 of ARR, you’ve created $10 of market cap from $2 invested. But as Rory cautioned, this only works when multiple hidden assumptions hold true:* The ARR is actually real (not inflated by aggressive accounting)* Net retention accounts for real churn (fast growth can hide massive customer losses)* Gross margins are sustainable (hyper-growth can mask deteriorating unit economics)* No massive capex requirements (true for most SaaS, decidedly not true for AI infrastructure companies)Jason Lemkin emphasized the existential risk everyone’s ignoring: “David Sacks coined the Burn Rate Multiple and I think when everything was the same in SaaS and B2B in 2021, it made a lot of sense. All the companies were the same… Then AI breaks it because we’ve never seen growth like this. But the margins are lower.”The team agreed that while burn multiples remain useful for comparing companies, they’re no longer sufficient as the primary valuation framework. As Rory put it: “We are not in Kansas. There’s a whole bunch of implied assumptions in there which is why even though we love those kind of metrics… we’ve actually come back to saying there’s a real advantage in seeing the GAAP revenue accounting also to make sure all the money is for lack of a better word just showing going up for real, right?”The Brutal Reality: Good Companies Can’t Get FundedHarry Stebbings raised what may be the most troubling question for founders: “I have many companies with good burn multiples and they are not getting love. They are not getting attention and they are going, ‘Harry, I don’t get it. I’ve been brought up to understand burn multiples, to understand growth, like what is going on?’”Rory’s response was unsparing: “There’s only two ways of pricing a deal. You price a deal on hope or you price a deal on the multiples. When you price a deal on hope and growth, you can lean in on anything, right? And you can get prices that quote unquote make no sense because the growth ultimately comes and it all pays off. Once you start valuing things on quote unquote the fundamentals today, right? Then you can value a public company because at 400 million, it’s not nothing. But… a $15 million revenue company that’s perfectly good and has reasonable growth is actually of zero value to a VC because we’re in the upside option game, right?”This crystallizes the existential crisis for non-AI-native companies: even with objectively strong metrics, if the path to a massive IPO isn’t crystal clear, VCs simply won’t engage. The bar isn’t just high—for certain categories, it functionally doesn’t exist.Jason was even more direct about the advice gap: “I hear too many folks, they’re like, ‘Oh, you’re triple triple double double or better. You’re you’re golden. Don’t worry, kids.’ And I think that’s terrible advice in 2025. Terrible advice.”He described portfolio companies at $15M ARR growing 100% with good burn ratios that can’t get funded unless they’re ultra-breakout AI stories. His advice? “If Scale wants to put in money at 250 on that deal… take that deal now.” Even if it means accepting a lower valuation than 2021-era thinking would suggest.Rory concurred: “Those companies, Jason, you’re right, is that… you should just get the deal done. Raise at a reasonable price. Continue to grow, but be capital efficient. Don’t get lost in your just your burn multiple. Focus on your cash. You know, if you’re right about your business, you’ll be right in the end.”The Kingmaker Effect: Why Proximity to Harvey or Abridge MattersHarry introduced a phenomenon that’s reshaping competitive dynamics: the power of King-Made Startups: “The mimetic—and this sounds obvious given the sheep plate analogies applied to venture—but it’s how concerned investors are by going against a kingmaker whether it’s Harvey or Abridge or any of the king-made companies. We have a couple of companies which are the second or the third and going against the kingmaker in the valley is the most unpopular thing in the world.”The strategy is working: “Founders listening, when you’re raising, raising to deter others from raising is a really working strategy right now.”Rory acknowledged the effect but with important caveats: “Do I believe that thinking exists in venture? Yes, I do. I don’t fully share it but I acknowledge that you know you have to factor it into your decision-making and your risk analysis… If your first customers are also VC-backed companies, then you get this dual loop. But look, the reality is… if you’re selling to oil and gas companies, they barely can tell the Sequoias from their KPs.”The takeaway: the kingmaker effect is real and powerful in Silicon Valley-centric markets, but less dispositive in traditional enterprise segments where buyers don’t track VC pedigree.Jason noted what makes AI different from previous cycles: “What is a little different in AI is that there are in many cases there’s not an established brand and there’s so much change and so much new budget and so much confusion that so many buyers are under pressure and have a desire to make a purchase… Being number one is so powerful when people know they want to do something—they’ve got to do an LLM for legal.”He cited the death match between Lovable and Replit as an example: “They’re both at nine figures in revenue. They won’t kill each other, right? But… you want to be that brand that nervous folks don’t know who to buy.”Valuation Insanity or Justified Vision? The $1 Trillion QuestionHarry posed the fundamental question: “Do you think we are near peak madness guys or do you think we’ll look back and laugh at ourselves for having this conversation given the might and the size of the markets that we’re entering?”Rory’s answer was unequivocal: “One of two things has to be true because that’s actually the interesting insight… If that massive transfer of labor doesn’t happen, then all these valuations are wrong by an order of magnitude. One of two things is going to happen in the next five or seven years, right? Either… you are going to see pretty profound productivity changes. You’re going to see companies like OpenAI hitting two to 300 billion in revenue super quickly. That’s option A. Option B… AI is still going to be wonderful, but you’re going to have a readjustment period that’s going to make your head hurt.”Jason worried about a different risk—loss ratios: “What I wonder is, are our loss ratios correct? If as venture VC firms, especially with larger funds… as long as we get it, as long as we’re cool, if 80% of our unicorns implode or blow up or make no money, more importantly, just don’t make money for venture… But I’m pretty sure we didn’t model it right in 2020 2021.”He continued: “It feels almost risk-free again like it did in 2021. It feels risk-free. Valuation and loss ratios both matter.”Rory distinguished between two types of fund-killing mistakes: “You can imagine people failing because their picking ratio was bad and they just put too much of the money in bad companies, right? Or you can also imagine people failing where all their companies were great, were good companies. But the prices were so bad that they only made a subpar return even though they only had a 30% loss ratio, right? So you both—I mean both things can—I mean the sad thing about venture is you actually have to get both things right to make money.”On whether this is sustainable, Rory offered a sobering forecast: “I personally think that the rate of adoption forecast over the next four or five that’s implicit in all these data center assumptions will in retrospect prove to be too optimistic.”Sam Altman’s Energy Crisis: Willing a Trillion Into ExistenceThe discussion turned to OpenAI’s staggering infrastructure requirements: planning to 125x energy capacity in 8 years, requiring more energy supply than India’s current total capacity.Jason framed the scale in visceral terms: “Just the little hundred billion that he’s doing with Nvidia… needs more power than all of New York City. So it will only be a couple years where the cities of the future don’t even have humans in them… Our country will be dotted with these Stargates that are larger than New York City with only hundreds of people working in them and the equivalent of millions of digital minds.”Harry expressed fundamental skepticism about the financing: “I feel very stupid honestly because I look at a trillion dollars required to fund data centers for Open AI alone. I’m like I don’t know where that money comes from… Sovereigns don’t have that. I mean, well, actually, funny enough, they do. I mean, they’d have to put it all in.”Jason offered a nuanced view: “I think Sam is just willing as much of this into existence as possible because of the future. And if we come up short, if it’s $400 billion or $600… we don’t have to buy all the GPUs. It would be okay if they had to last 6 years instead of 3 years or whatever the depreciation schedule is.”Rory distinguished between directional vision and operational reality: “The sheer act of willing something into existence like this has been just amazing… Whatever the prize is for being the best company in AI, OpenAI is going to get that prize, right? And that’s his job. And he’s doing it better than any other CEO of this decade. That’s true. It is also equally true… just because a CEO says we’re going to treble next year doesn’t mean we should buy real estate and hire people as if we’re going to treble.”His concern: “Five years from now… we said, ‘Oh, I get it. OpenAI is still the best company on the planet for AI.’ Its growth rate has slowed to a shockingly small 50 60%. It’s freaking amazing. It’s doing 30 billion dollars growing at 50%. It’s astonishing. But maybe they don’t need a trillion dollars of capex this week.”Zuck’s AI Crisis: When Even Loyalists Lose FaithHarry made a rare public admission about his largest public holding: “I always thought that Zuck earned the right to do the next thing. He earned the right to do the next thing. And I have to say my faith in Meta’s AI strategy has just dwindled and dwindled and dwindled… I’m looking at Alex Wang. I’m looking at the treatment of Yann LeCun. I’m looking at how they structure teams and I’m going this is not wellrun. OpenAI, Anthropic are coming for you.”Jason was characteristically blunt: “He is not the communicator that Sam Altman is… If I don’t understand where the hell he’s going and Harry doesn’t understand… if we don’t understand he’s not one of the great communicators at the moment… He’s a crappy communicator right because none of us—I’m not saying his AI strategy isn’t S tier but I don’t think any of us understand that not for the life of us where the hell it’s going.”Rory offered both defense and critique: “You could also say it’s just a harder bet… The bet for the last 15 years was riding this brilliant invention that you had in 2003 and just optimizing it. And that’s hard but it’s a lot easier than now you got to invent a whole new thing a second time.”He continued with the existential framing: “The big picture comment is if you spend two hours a day on ChatGPT that’s two hours a day that you are not spending on Facebook and we live and die on our attention. So we’re just going to make s**t until somehow we get people to come back and play with us.”The panel agreed that while Zuck was “clear just this week he’d rather burn the 20 billion in operating income and fail than become irrelevant,” the specific strategy remains opaque even to sophisticated investors.Rory’s prediction: “If there was a Couch bet… that you know some version of this AI strategy will not produce meaningful revenue despite a $20 billion burn and will look more like Meta VR and less like Instagram and WhatsApp… I’d take that bet.”The Great Consolidation: Fivetran + dbt Labs and What It MeansThe panel discussed unicorn Fivetran’s talks to acquire unicorn dbt Labs as a template for the massive consolidation wave that must happen with 6,700 unicorns and only 15 IPOs year-to-date.Rory framed the math: “We’ve processed 15 out the IPO gate year to date. So 20 for the year. That implies we got 30 years of this stuff to get through. What they’ve done… every time two companies combine… it takes two midsize companies… and makes a bigger… This is part of the job venture is going to have to do to whip their portfolios into shape to be IPOable.”Jason identified the key enabler: “The fact that Andreessen is the lead or close to it in both deals makes it much simpler on many levels… If Andreessen owns 20% of Fivetran and 20% of dbt and you combine them, it does kind of suck when you own 20% of a portfolio company, combine it with another leader… and now I own 8% after the deal.”But he emphasized why it still makes sense: “20% of something that’s not going public is not nearly as interesting as 8% of something that is going public, right? And if you believe that it’s not a continuum of value… but rather it’s like electron states, there’s just a gap and then you got to go to the next state.”Rory stressed the risk: “The fatal mistake that always scares me is not the delusion… The thing that scares me is you go from a decent deal that’s wellrun where you know everything about it to merging with something else and then the combined entity screws it up. That to me is the really shitty outcome where you took your 20% bet and turned it into 8% of a disaster.”The PE Reckoning: Why Tech Private Equity’s Model Is Under ThreatHarry asked whether tech PE firms are questioning their business model: “When they see the multiples that you can get on the money that’s being moved by your Kushner and your big firms, combined by the increased loss ratio that will happen from an increasingly volatile new AI world. Are you suddenly going, ‘S**t, I’m not getting paid for the risk that I’m taking on the multiple on the upside given the displacement on the downside.’”Jason identified the core problem: “These products didn’t change from about 2008 until 2023. They’re the same products… It wasn’t just that we had high NRR which was the spreadsheet glue for the PE model. It was the fact… that the products can’t be static and AI is the accelerant there… The rate of change is unprecedented in business software. Unprecedented.”He gave the example of his first venture investment, Pipedrive: “It took him four years to launch a mobile app… Could you imagine today waiting four years to launch your AI co-pilot? Like you’re dead in the water.”Rory expanded on the implications: “There was 15 years where we made the same product and you didn’t have to think that much about product direction at the macro level… You look at Salesforce 2002 and 2022, it’s the same thing, right? And that’s now changed… On the PE side, you’re exactly right. You look at that and you go, I might get away with next 10 years making the same thing… And so maybe I’m entering into some kind of tech risk that I’ve never before internalized. A new tech risk.”Jason added the seat model crisis: “Now that we’re running 12 AI agents, you know, we only need two seats of Salesforce. Because you don’t have the people. We just don’t… It makes the PE model worse and our jobs harder. It just literally… makes it even tougher for PE to buy these seat models.”Key Takeaways* Burn multiples are useful but insufficient: AI companies break traditional SaaS assumptions around gross margins, churn visibility, and capex requirements. VCs must look at GAAP revenue and absolute cash positions, not just efficiency ratios.* Non-AI-native companies face a binary funding environment: Even with triple-triple-double-double growth and good burn multiples, companies perceived as “legacy” struggle to raise. Founders should take reasonable deals now rather than optimizing for price.* The kingmaker effect is real but not dispositive: In Valley-centric markets, being the #2 or #3 behind a heavily-funded leader significantly impacts fundraising. But in traditional enterprise segments, customer buying decisions care less about VC pedigree.* Massive consolidation is inevitable: With 6,700 unicorns and 15 IPOs year-to-date, the math demands hundreds of mergers. Firms like Andreessen that own both sides of deals have structural advantages in making these combinations happen.* Tech PE’s classic SaaS funding model faces existential threats: The shift from static products (unchanged for 10+ years) to rapidly evolving AI-powered solutions introduces unprecedented tech risk. Meanwhile, AI agents reducing seat-based revenue attacks the core cash flow assumptions.* OpenAI’s ambitions may be both directionally correct and operationally overstated: Sam Altman is “willing a trillion into existence,” but even if the vision is right, the actual capital deployed and timeline may be far more modest than the rhetoric suggests.* Meta’s AI strategy lacks clarity even to sophisticated investors: Despite a $20B annual commitment, investors including long-term bulls struggle to understand the strategic endgame beyond “avoid irrelevance.”* Loss ratios matter as much as entry valuations: Even if VCs pick the right companies, paying prices that require 100x returns to generate acceptable fund returns means most funds will fail even with good picking.Quotable MomentsOn burn multiples in the AI era:* “A $15 million revenue company that’s perfectly good and has reasonable growth is actually of zero value to a VC cuz we’re in the upside option game.” — Rory O’DriscollOn outdated advice:* “I hear too many folks, they’re like, ‘Oh, you’re triple triple double double or better. You’re golden. Don’t worry, kids.’ And I think that’s terrible advice in 2025. Terrible advice.” — Jason LemkinOn the kingmaker phenomenon:* “Going against the kingmaker in the valley is the most unpopular thing in the world. Founders listening, when you’re raising, raising to deter others from raising is a really working strategy right now.” — Harry StebbingsOn OpenAI’s prize:* “Whatever the prize is for being the best company in AI, OpenAI is going to get that prize.” — Rory O’DriscollOn Sam Altman’s energy vision:* “Our country will be dotted with these Stargates that are larger than New York City with only hundreds of people working in them and the equivalent of billions of digital minds.” — Jason LemkinOn financing the impossible:* “I feel very stupid honestly because I look at a trillion dollars required to fund data centers for Open AI alone. I’m like I don’t know where that money comes from.” — Harry StebbingsOn Meta’s communication problem:* “He’s a crappy communicator right because none of us—I’m not saying his AI strategy isn’t S tier but I don’t think any of us understand that not for the life of us where the hell it’s going.” — Jason LemkinOn tech risk in PE:* “Maybe I’m entering into some kind of tech risk that I’ve never before internalized. A new tech risk.” — Rory O’DriscollOn consolidation math:* “20% of something that’s not going public is not nearly as interesting as 8% of something that is going public.” — Jason LemkinThanks for reading The Secrets To Scaling in The Age of AI! Subscribe for free to receive new posts and support my work. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

Tomasz Tunguz of Theory Ventures brought together a strong panel at SaaStr Annual and AI Summit for a deep dive on enterprise partnerships. An Enterprise Partnerships Bootcamp: How to Land, Scale, and Win with Linear, Omni, Theory Ventures, and Vesey Ventures.* Christina Cordova – Chief Operating Officer at Linear, the purpose-built tool for planning and building products. Previously spent 7.5 years at Stripe where she started their partnerships organization, and later led partnerships and platform initiatives at Notion.* Colin Zima – CEO of Omni, rebuilding BI and analytics to create a tool accessible to both developers and end users with enterprise readiness and AI agility. Previously spent nearly 10 years at Looker, bringing deep data ecosystem expertise to his current venture.* Julia Huang – Founding Partner at Vesey Ventures, a fintech fund based in New York and Israel. Specializes in brokering partnerships between portfolio companies and financial incumbents, with deep expertise in enterprise sales motions within regulated industries.* Tomasz Tunguz – Founder and General Partner at Theory Ventures, focusing on early-stage enterprise software investments with particular emphasis on go-to-market strategy and partnership development.Top Take-Aways:Christina Cordova (Linear): Start with partnerships that are instrumental to your product experience first, not distribution. Build credibility by talking to 10+ users within an enterprise before approaching executives.Colin Zima (Omni): Partner with companies at your growth stage rather than chasing Fortune 50 whales early. Strategic investments from key ecosystem players can bootstrap credibility faster than organic growth.Julia Huang (Vesey Ventures): In enterprise partnerships, you’re controlling for reputational risk, not financial loss. Find the sponsor who can tell your story back to you—they’re your true advocate.Tomasz Tunguz (Theory Ventures): The cost of building integrations has become trivial with AI. Companies should have 25-100 integrations by Series A, not 5-10 like in the past decade.The Partnership vs. Sales Decision: Where to Start and WhyThe fundamental question facing early-stage B2B companies isn’t whether to pursue partnerships or direct sales—it’s understanding when partnerships become instrumental to your product experience and business growth. Linear’s approach exemplifies this strategic thinking.“For us on the partnership side, we really started with partnerships that we felt were instrumental to the product experience first and foremost,” explains Christina Cordova. “At a certain point we were a very small company. We viewed some integrations as hyper-critical to the core product—things like integrations with Slack and GitHub.”Quotable Moment – Christina Cordova: “We were able to have 10 conversations for a very large account and then we go back to the CTO and say, ‘Actually, we’ve talked to 10 people on your team and here’s what they say about the products that you’re using today.’ And then they’re kind of like, ‘Oh, you know more about my team and my org than I do.’”This product-first approach to partnerships reveals a crucial insight: the most successful early partnerships aren’t just distribution plays—they’re core to delivering your value proposition. Linear recognized that without seamless integrations to the developer tool stack, their product experience would be fundamentally incomplete.The Evolution of Partnership Strategy: From Product to PlatformOmni’s partnership evolution illustrates how dependency relationships in the enterprise ecosystem have fundamentally shifted over the past decade. Colin Zema’s perspective, shaped by his experience building partnerships at both Looker and Omni, reveals how the technology stack has matured.“We can’t deliver our product without a whole bunch of other products tightly coupled to us,” Zema notes. “We need to get data into a database. We need a database to run the queries and then we need to present them out to the user. At every level to be successful, we have these contingencies on these other tools and they need to want to work with us also.”Quotable Moment – Colin Zema: “I can go call all of you in the audience and ask you to come try my tool. It’s a lot more successful if Snowflake is doing that alongside us and if DataBricks is doing that alongside us.”But the partnership landscape has evolved dramatically. When Looker was scaling, companies often had to sell both ETL solutions and their core product to be successful. Today, that infrastructure is assumed. “Everyone sets up a database, buys a database, sets up Fivetran, and now how we partner with them is more important than whether they exist,” Zema explains.This shift has profound implications for partnership strategy. Modern partnerships are less about filling infrastructure gaps and more about creating seamless experiences within established ecosystems.The Credibility Framework: Building Trust at Enterprise ScaleEstablishing credibility emerges as the central challenge—and opportunity—in enterprise partnerships. Julia Huang’s experience working with financial incumbents provides a masterclass in managing reputational risk.“The risk that we were controlling for isn’t financial loss. It was reputational risk,” Huang explains. “If their reputational brand and image gets tarnished, then they don’t have the ability to sell their customers anything else.”Quotable Moment – Julia Huang: “My practical advice on finding your sponsor is the person who can tell your story back to you in that meeting. Somebody who’s starting to get it—the wheels are turning. That person will at a minimum give you feedback, the true feedback of where you sit within the product and how long it’s going to take.”This insight fundamentally changes how startups should approach enterprise partnerships. Success isn’t just about proving product-market fit—it’s about demonstrating that your company can enhance, rather than threaten, your partner’s reputation with their customers.The credibility challenge manifests differently across organizational levels. When selling to individual contributors, you’re proving quarterly value. When selling to VPs, you’re demonstrating annual impact. When engaging C-level executives, you’re articulating a three-year vision that aligns with their strategic objectives.The API-First Partnership PlaybookLinear’s approach to building their partnership ecosystem offers a replicable framework for developer-focused products. They started with a GraphQL API and “shipped the API and didn’t do a whole lot after that to bring on partners outside of just having really good technology.”This hands-off approach generated over 150 integrations organically. But Linear’s strategy evolved with market conditions, particularly around AI partnerships. “More recently, we’ve thought more about partnerships in the context of AI in particular because it’s such a fast-moving space,” Cordova explains. “There we’re going much more the partnership route where we’re going to all the companies and spending time with those companies.”The strategic shift reveals when to move from passive platform provision to active partnership development: when the technology landscape is moving so quickly that organic adoption can’t keep pace with market opportunities.Strategic Investment as Partnership AccelerantThe relationship between strategic investment and partnership success has become increasingly important. Omni’s experience securing investment from both Snowflake and DataBricks illustrates how financial alignment can accelerate go-to-market execution.“Both of those in isolation have been huge legitimizers for us as a company,” Zema notes. “It gives us opportunity to build more product over time and lets us bootstrap credibility with our customer base.”But the investment decision requires careful consideration. “We’re way more concentrated in terms of our partnership,” Zema explains. “The databases are the most important to us. So it’s Snowflake, BigQuery, RedShift, DataBricks… We can’t move the needle for Snowflake’s earnings, but Looker did. And so we want them to trace that curve out and see that forward.”The Timing Revolution: When to Hire Your First BD PersonTraditional wisdom suggested waiting until $5-10M ARR before engaging seriously in partnerships. Tomasz Tunguz challenges this conventional thinking: “I think that’s really changed. One of the earliest employees at Looker, Kenan was BD. He was responsible for 10 to 15% of bookings from the very first.”The catalyst for this change? Technology has dramatically reduced integration costs. “The cost to produce an integration, an API integration, is really trivial,” Tunguz explains. “One of our companies had a hack day where in 90 minutes an engineer spun up an integrations factory.”Quotable Moment – Tomasz Tunguz: “You kind of expect as a Series A investor a company having 25, 50, maybe even 100 integrations by the time of the Series A because it’s so simple. Whereas 10 years ago, that would be almost unimaginable because the engineering work required to do that would have taken years or hundreds of thousands of person-hours.”This technological shift means companies can now reasonably be expected to have 25, 50, maybe even 100 integrations by Series A—a scale that would have been unimaginable a decade ago.Christina’s hiring philosophy provides practical guidance: “When you’ve tested it with your current team, what are the kinds of partnerships, what do they look like? Once you know you want to do more of these and it’s going to be really a full-time job, then it makes sense to hire someone to do these in a dedicated fashion.”Channel Strategy: Bottom-Up vs. Top-Down ApproachesThe choice between bottom-up and top-down partnership approaches depends fundamentally on your product’s natural adoption pattern. Products that win on user experience often benefit from grassroots adoption, while products that solve C-level problems may require executive-first approaches.Linear’s bottom-up strategy leverages product enthusiasm to build executive credibility. “We’re able to have 10 conversations for a very large account and then we go back to the CTO and say, ‘Actually, we’ve talked to 10 people on your team and here’s what they say about the products that you’re using today,’” Cordova explains.This approach works because it builds internal advocacy before asking for budget authority. The CTO realizes, “You know more about my team and my org than I do,” creating credibility that transcends traditional sales processes.The Enterprise Breakthrough: Acquisition-Driven ExpansionSome of the most successful enterprise partnerships emerge through acquisition scenarios. Both Linear and Omni have experienced rapid expansion when their startup customers get acquired by larger enterprises.“We started off with this company of like 50 people. It was ultimately acquired by a Fortune 20 company and that Fortune 20 company started using us,” Cordova shares. “We were able to take that one customer and expand it to like a 500 seat account very quickly.”This pattern reveals a crucial insight: startup CEOs often retain significant influence post-acquisition and can advocate for tools they’ve grown to depend on. It’s a channel strategy that requires no additional sales effort—just product excellence that creates passionate advocates.The AI Partnership AccelerationThe current AI wave has created new urgency around partnership development. Both Linear and Omni are adapting their partnership strategies to capture AI-driven opportunities.Linear is building new APIs specifically for AI agents and taking a much more proactive approach with AI partnerships. “There’s now this weird discretionary pool of cash that people can use for tools if you check the AI box,” Cordova notes.This budget reality means partnerships that can credibly position AI capabilities have accelerated paths to enterprise adoption. The challenge is ensuring the AI positioning represents genuine value rather than checkbox marketing.Measuring Partnership Success: Beyond Revenue AttributionTraditional partnership metrics often focus on directly attributable revenue, but the most successful partnerships create value that’s difficult to measure directly. Brand credibility, ecosystem positioning, and customer trust building may be more valuable than immediate revenue in early-stage partnerships.Colin’s experience illustrates this: “It’s a lot more successful if Snowflake is doing that alongside us and if DataBricks is doing that alongside us.” The value isn’t just in the leads generated—it’s in the credibility transfer that happens when established players validate your solution.The Partnership Portfolio StrategyAs companies scale, partnership strategy must become more portfolio-driven. Not every partnership will generate immediate results, but a balanced approach can create multiple paths to success.“I think as long as you place the right ones, it can be very fruitful,” Cordova explains about Linear’s approach to partnering with both established AI companies and emerging startups. “Some of those bets aren’t going to work out, but I think as long as you place the right ones, it can be very fruitful.”This portfolio thinking requires discipline about resource allocation and clear criteria for partnership investment levels.The Procurement Reality CheckEnterprise partnerships ultimately face the procurement reality of large organizations. Colin’s experience provides sobering perspective: “You realize that the conversations are actually completely different. The person may not have any idea what your company does. They have no ability to buy.”This reality requires partnership professionals to develop multiple messaging frameworks for different organizational levels and decision-making authorities. The technical value proposition that resonates with practitioners may be irrelevant to budget holders focused on strategic outcomes.Building Partnership Infrastructure for ScaleSuccessful partnership programs require infrastructure that can scale beyond individual relationships. APIs, documentation, partner portals, and co-marketing resources become essential as partnership volume increases.Linear’s evolution from building individual integrations to providing platform capabilities illustrates this maturation. The infrastructure investment pays dividends when partners can self-serve many of their integration needs.The Strategic Investment Decision FrameworkWhen considering strategic investments from partners, companies must balance accelerated credibility against potential constraints on future partnerships. The decision requires careful analysis of how strategic alignment affects future partnership optionality.Omni’s concentrated approach—focusing strategic investment on their most critical database partners—provides a model for how to think about these tradeoffs. “You can’t do a hundred of them either,” Zema notes.Top Mistakes the Speakers Made (And What We Can Learn)Christina Cordova’s Mistake: Being too hesitant to hire dedicated BD resources. “I’m always very hesitant to make that hire because I’ve seen it so many times how it can so easily go wrong.” While caution is warranted, this conservative approach may have delayed partnership scaling at Linear compared to more aggressive competitors.Colin Zima’s Mistake: Admitting to poor executive communication. “I did a conversation last week with a CIO of a Fortune 50 and at the end he was pretty much like I had no idea what you were talking about.” This reveals insufficient preparation for senior-level conversations and highlights the critical importance of adapting technical messaging for different audiences.Julia Huang’s Mistake: Allowing strategic partnerships to drag on too long without clear milestones. She describes partnerships with banks that “take years to come into fruition” and mentions going through reorgs “three times or four times” during one strategic partnership. Better partnership governance and clearer exit criteria could prevent resource drain.Tomasz Tunguz’s Mistake: Overestimating the speed of integration development. While he correctly identifies that integration costs have dropped dramatically, his “90 minutes” integration factory example may set unrealistic expectations for quality, scalable integrations that require proper testing, documentation, and ongoing maintenance. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

This week: * How the new Replit v3 is the Future: Agents Managing Agents, For Real* How our 4th AI “BDR” helps close deals 24x7, and is much better than a human at it* How our new SaaStr.ai Startup Valuation Calculator processed 300,000+ AI startup valuations in less than 30 days. And how disruptive (and just plain cool) the new Replit v3 is. At SaaStr, we’ve gone from having essentially zero AI agents at the start of 2025 to now having over 12 AI agents in production. We have AI SDRs and BDRs handling both inbound and outbound, a digital assistant answering 150,000+ chats on our websites, and we’ve built AI tools that have processed over 300,000 startup valuations and graded nearly 1,000 VC pitch decks - all in less than 30 days.I think we’ve learned something here about how to effectively build, deploy, and scale AI agents in a real B2B business. And more importantly, what breaks, what doesn’t scale, and where the real ROI lives.Here’s what we learned across three major areas: the future of AI development platforms (Replit V3), why you absolutely need AI BDRs now, and how we built tools that generated real revenue without a development team.Replit V3: The First Glimpse of AI Managing AIThe Big Question: Can you build real B2B applications without a team of developers? The answer is increasingly yes, but it’s more complicated than the marketing suggests.I started this journey a couple months ago looking at the leaders: Replit, Lovable, Bolt, and hot newcomers like Wix Studio and Base 44 (which may have 10% market share already). Many are doing nine figures in revenue very quickly.Why I Chose Replit: When I asked Twitter which platform to pick, everyone said they were similar — at time, a few months back. But Replit was the only one where you could go end-to-end - build, test, prototype, and push to production without configuring databases, moving to different hosting, or dealing with infrastructure headaches. The white-labeled Neon database made it plug-and-play with other tools.Agents Managing Agents: It’s Already HereBut here’s what blew my mind with V3: Replit now has agents that manage other agents.Most of us are struggling to get one AI SDR working, let alone having AI manage 20 agents. But Replit can do it today. When I hit a complex problem building our pitch deck grader, Replit autonomously brought in:* An architect for really tough problems* Specialists for specific issues* Senior and junior agents with different capabilitiesI watched these agents debate each other (in English) for almost 3 hours while conducting a security audit. They went through every line of code, every function, every page, debating how much to lock down our SaaStr AI app for security.The Reality Check: When Even The Best AI Perhaps Is Too PowerfulThe agents did incredible work autonomously, but in some ways, they did go too far. By the time they finished, the app was so locked down for security that almost nothing worked. You couldn’t upload PDFs, couldn’t upload anything, analytics were blocked - it was all blocked as security risks.The result: I spent over 10 hours undoing the security audit, retesting every page and link. Everything interactive had been locked down to be “super secure” but became in many cases non-functional.The Learning: This level of autonomous capability is incredibly cool, but it’s different. Many Replit users revolted when V3 launched because:* Costs went up (smarter AI uses more tokens)* It’s slower (smarter AI takes more time)* No option to stay on V2* Required learning new workflows and autonomy settingsThe Business Process Change Problem: We’re moving at AI speed because we have to - competition is fierce. But as we go mainstream, we’re colliding with the reality that most users only want so much business process change. V3 is incredibly powerful, and it’s so cool. But if V2 was good enough, many users didn’t want to invest the learning curve.This is a pattern we’ll see across all AI tools as they rapidly evolve.Why You Need AI BDRs Now (And How to Train Them Right)We now have four AI agents handling different parts of our revenue funnel:Three AI SDRs (using Artisan):* Outbound to potential sponsors and partners for events* Outbound to prospects for event attendance* Recirculation to previous attendeesOne AI BDR (using Qualified):* Manages inbound on SaaStrAnnual.com and SaaStrLondon.com* Integrates with all our Salesforce and Marketo data from the past decade* Qualifies prospects without them feeling like they’re being qualified* Sets up meetings in real-time with our sales teamWhat Makes This AI BDR DifferentGo to SaaStrAnnual.com or SaaStrLondon.com and try the chat. Superficially, it looks like every other chat bubble (Intercom, Finn, Drift - they all look the same). But here’s what’s clever:It knows everything about you already. If you’ve sponsored before or been to a SaaStr event over the last decade (50,000+ people have), we already know about you. If you’re a potential sponsor, it’ll tell you about similar companies that sponsored last year and their results. If another AI SDR company visits, it’ll mention how Artisan got $500,000 of pipeline from last year’s event.It does three things that are game-changing for tiny teams:* Qualifies prospects better than humans - without the soul-crushing experience of talking to some junior person who knows nothing and is just deciding if you’re worth an AE’s time* Solves problems and answers questions using a decade of integrated data* Automatically sets up meetings with the right sales rep in real-timeIt can even create new opportunities. Someone reached out wanting to bring 30 CEOs to SaaStr AI London. The AI created a whole package and suggested the idea to us.The Training Reality: It’s Not What You ThinkYou absolutely must train these agents relentlessly. But here’s what we learned that most people get wrong:You don’t need as much data as you think upfront. Our first agent was trained on 20 million words of SaaStr content. But when the CEO of Deli compared our agent to Brian Halligan’s, Lenny Rachitsky’s, and Keith Rabois’s agents (all on the same platform), we learned that massive historical content isn’t that valuable.What you actually need:* A solid base of good content (your website, sales pitch, collateral, reference customers)* Relentless daily iteration and AB testing* Watching every interaction in real-time* Proofing and correcting anything wrong* Constantly adding improvementsSet and forget does not work. You have to relentlessly iterate on these AI platforms.Two Examples of Terrible AI “Non-Training” (And Why You Can Do Better)Example 1: Brex’s AI Support DisasterBrex kept sending alarming emails to us demanding we “immediately deposit $5 million at Brex” to maintain credit limits. When I tried their AI support (using Decagon), it wouldn’t even let me chat with AI or humans - it just closed the window.The Fix: This needed zero complex training. Just one human sampling issues and uploading content explaining: “That email sounds alarming, but you don’t need to deposit $5 million. Here’s how Brex credit actually works.” Thousands of people got this email. No excuse for not training the AI.Example 2: Generic Zendesk AI SupportWhen I had a specific infrastructure issue, Replit’s Zendesk AI kept asking for screenshots and giving generic answers. It wasn’t trained on the exact issue and couldn’t solve something that would take minutes with proper training. It wasn’t that hard to fix — a great human actually help me fix it in minutes. But the AI was utterly lost, and kept asking me basic questions again and again. All day long. The Lesson: Unlike humans, AI agents won’t keep making the same mistakes once properly trained. But you have to do the work upfront and continuously.Building Revenue-Generating Tools: 300,000 Valuations and 1,000 Pitch DecksAt SaaStr.ai, we built two tools that generated real user engagement:AI Startup Valuation Calculator: 300,000+ free valuations in 30 days VC Pitch Deck Analyzer: Almost 1,000+ founders uploaded decks in 10 daysBoth are free, fast, simple, and don’t require registration. But here’s the key learning for anyone building complex AI applications:The Secret: Offload the ComplexityMy first vibe coded app failed because I tried to build matching algorithms for 10,000 SaaStr attendees (10,000 x 10,000 potential matches) in real-time within Replit. Way too complex. But my next 7 got into production. For the valuation calculator:* I took thousands of funding rounds and data points* Distilled them in Claude into four simple tables instead of asking Replit to crunch all the data each time* Simplified 5,000+ funding rounds into manageable datasets* Let OpenAI API handle the calculationsFor the pitch deck analyzer:* Combined learnings from 800+ VCs who spoke at SaaStr Annual* Simplified complex analysis into a manageable prompt for Anthropic API* Ran 100+ test decks (including recent funded SaaStr Fund companies) to tune the weights* Let Anthropic do 90% of the heavy liftingThe Pattern: Have AI do as much of the complex work as possible, but pre-process and simplify the inputs.The Two-Domain Problem: A Glimpse of What’s ComingWe now run SaaStr.com (WordPress, 35 million SEO views/year) and SaaStr.ai (Replit-built, dynamic, modern) simultaneously. Why?The Dilemma: Do you want state-of-the-art dynamic platforms that ca.n do things WordPress can’t, or do you want proven SEO and stability?Moving off SaaStr.com’s domain, static URLs, and content to a new platform is super risky. The new SaaStr.ai is shiny and cool, but doesn’t have the same SEO benefits.The Reality: We’re probably not the only ones dealing with this. Many companies will face this choice in coming years. Exporting AI code to run in WordPress isn’t elegant - it’s slow, clunky, and you lose the ability to rapidly iterate and add features.We’re at the bleeding edge here, with about 30,000 monthly users on SaaStr.ai (10-20% of our traffic), but it’s a preview of infrastructure decisions every B2B company will face.The Bottom Line: What Actually WorksFor AI Development:* For non-developers, choose platforms that let you go end-to-end without infrastructure headaches* Agents managing agents is here today and super powerful, but prepare for cleanup time* Business process change resistance is real - factor it into rolloutsFor AI Revenue Operations:* You need AI BDRs now - your competitors will have them soon* Training is everything, but you need less data than you think upfront* Daily iteration and monitoring is non-negotiable* The experience should feel human, not like talking to a botFor Building AI Tools:* Offload complexity to AI APIs rather than building everything in your platform* Pre-process and simplify inputs rather than asking AI to handle raw complexity* Focus on solving real problems people will pay for or engage with repeatedlyThe Meta-Learning: We now have more AIs than humans at SaaStr. The technology works, but success comes down to relentless training, iteration, and understanding that AI agents are tools that amplify good processes - they don’t fix bad ones.Try our tools at SaaStr.ai (hit “AI VC” at the top), and let me know what you build with your own AI agents.What AI agents are you putting into production? What’s working and what’s breaking? Share your experiences in the comments.Thanks for reading The Secrets To Scaling in The Age of AI! Subscribe for free to receive new posts and support my work. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

Harry, Rory and Jason are back!We’re witnessing an unprecedented capital concentration in AI with NVIDIA’s $100B OpenAI investment creating a fascinating circular money machine, while new H-1B visa fees threaten startup talent acquisition and the venture funding landscape shifts dramatically toward mega-rounds for a tiny number of companies. The era of “founder friendly” has become somewhat hollow rhetoric, and traditional B2B growth metrics like “triple triple double double” are becoming irrelevant as the market polarizes between AI unicorns and fundamentals-driven businesses.Key Numbers That Matter:* 75% of 2025 VC dollars went to just 19 companies* NVIDIA’s $4.5T market cap relies on only 6 customers for 83% of revenue* New H-1B visa fees of $100K will impact 440,000 annual applications* Navan filing for IPO at $8B valuation with $613M revenue, 32% growthThe $100B AI Money Machine: When Six Customers Drive a $4.5T Market CapNVIDIA’s massive investment in OpenAI represents more than just capital deployment—it’s the creation of what could be an infinite money printing loop. OpenAI commits $300B to Oracle, Oracle buys NVIDIA chips, and NVIDIA invests back into OpenAI. As one observer noted: “Sam’s gonna get to make the bet he wants to make which is apply infinite amount of capital and see how long these scaling laws last.”The most striking aspect? NVIDIA, now the world’s largest company by market cap at $4.5 trillion, has only six meaningful customers accounting for 83% of revenue. Compare this to Apple’s 2 billion customers or Microsoft’s hundreds of thousands of enterprise clients. It’s “this really weird dynamic where you’ve got this company with only six customers, but the good news is all six of them are determined to spend themselves into oblivion to win the prize.”The Scaling Laws GambleSam Altman’s recent comments suggest this is just the beginning: “We need three orders of magnitude more compute than this.” The market is essentially allowing OpenAI to test whether massive capital can break through current AI limitations. Whether the marginal $300 billion will earn a return on capital remains questionable, but as Rory put it: “We will find out because no one’s going to call timeout along the way.”The H-1B Shock: $100K Fee Creates New Startup RealityThe new $100,000 fee for H-1B visas represents a significant shift for the startup ecosystem. With 440,000 applications generating $19-120 billion in GDP annually, this policy change will have material impact on early-stage companies.“Anyone that has been doing this for a while that isn’t just three kids working 24/7 in SF has had H-1B folks on their team,” noted one investor. “My first startup wouldn’t have been possible without H-1B. I had two on my first team of 10.” notes Jason.While larger tech companies will simply absorb the cost, startups face a more complex reality. The workaround? Most founders are now pursuing O-1 visas, though these come with their own complications and ongoing maintenance requirements.Venture Capital’s Great Concentration: 75% of Dollars to 19 CompaniesThe venture landscape has fundamentally shifted. In 2025, 75% of VC dollars went to just 19 companies—a stunning concentration that reflects the bifurcation between traditional venture and ultra-late-stage private public investing.“What’s really happened is on top of that business has emerged this totally separate business called ultra late stage,” explains Harry. “It just means there’s another business that you can choose to be in or not that exists one or two orders of magnitude above you in the valuation world.”The Death of “Triple Triple Double Double”?Traditional SaaS growth metrics are becoming obsolete for many companies. While “triple triple double double” (3x-3x-2x-2x growth pattern) remains relevant at early stages, it’s insufficient for late-stage rounds in today’s market.“There’s only so many folks growing beyond triple triple double double at 50 to 100 million—VCs will take the meeting,” notes Jason. But for companies in traditionally unloved verticals or without AI narratives, even strong growth isn’t guaranteeing funding notes Harry.The Founder Friendly FacadeThe concept of “founder friendly” has become meaningless in today’s hyper-competitive environment. “Founder friendly has become b******t,” argue both Harry and Jason. “Any hot AI deal, there is no diligence provided, nor is any done. It’s just done on Saturday.”Real founder-friendly behavior shows up in difficult times: “Founder friendly is writing the check when no one else does. That’s founder friendly. Founder friendly is when no one else is there at the board meeting anymore and you’re there and you still have a W on the other side of it.”Market Concentration Creates New DynamicsThe concentration extends beyond just funding to data labeling and infrastructure. Multiple data labeling companies report that the same two AI giants make up 55% of their revenue across all providers. This concentration creates a unique dynamic where “none of those six customers is trying to optimize their cost basis. They’re just trying to build as fast as they can.”IPO Strategy in a Concentrated WorldCompanies like Navan (filing at $8B with $613M revenue, 32% growth) are timing their IPOs strategically. Rather than waiting for profitability, they’re moving quickly to avoid being the “third player” in their category after competitors like Brex and Ramp potentially go public.“When people perceive it as a direct comp, even if it isn’t, it becomes troubling to get out because every public investor says, ‘I already have Brex and Ramp. Why would I buy you?’” notes Rory.The New Venture Reality CheckFor investors, the current environment requires recalibrating expectations. Jason admits: “I’m 0% cash, just like 2008. I remember feeling I could literally not have enough cash to fix the roof on my house—nothing was more fun than selling my stock at a 70% loss to fix that roof.”The frothiness is evident: LPs are now bragging about returns on LinkedIn, historically a sign of market peaks. Yet for companies with genuine traction, funding remains available. The key is having clarity on realistic next-round pricing rather than extrapolating long-term returns.Portfolio Concentration LessonsSuccess breeds more risk tolerance. “Having an early success is highly correlated with future success. Partly you get the referral effect, but partly I think it’s that you just have the stomach to roll the dice and you get braver.” notes Rory.This explains why firms like Sequoia can hold positions like Klarna while emerging managers face different portfolio pressures around concentration limits.Looking Ahead: The End of 2021 ValuationsThere’s growing consensus that 2021 valuations should be written off entirely. “I think we can’t talk about 2021 valuations anymore. It’s time to just flush them down the toilet,” suggests Jason proposing a deadline: “You’ve got 90 days left to caveat and complain about your 2021 valuations. And on January 1, 2026, no one is allowed to talk about their 2021 valuations.”Companies like Notion, now at $500M ARR with re–\acceleration, show that B2B companies can recover by embracing AI while maintaining their core business. The path forward isn’t becoming “something totally different” but rather intelligent integration of AI capabilities.The Monopoly QuestionOpenAI’s consumer dominance raises antitrust questions, but it’s complicated by their massive subsidization of users. “Never have there been a less successful monopoly than ChatGPT because they’re subsidizing the consumer surplus delivered by ChatGPT in the tens of billions of dollars” notes Rory.The real monopoly question may be more relevant to NVIDIA, which has true market share dominance in AI chips, making their equity investment in OpenAI particularly strategic.Key Takeaways* Capital concentration is unprecedented: 75% of 2025 VC dollars went to just 19 companies, creating a new tier of ultra-late-stage private investing distinct from traditional venture* H-1B changes will impact startups materially: The $100K fee creates real barriers for early-stage companies, though workarounds and for founders use of O-1 visas are and will emerge* “Founder friendly” has become meaningless: In hot AI deals, there’s no diligence and term sheets are signed on Saturdays, making the concept hollow* Traditional SaaS metrics are obsolete: “Triple triple double double” no longer guarantees late-stage funding except for the hottest AI companies* Timing matters for IPOs: Companies are racing to go public before competitors in their category, as being third creates comparison problems* 2021 valuations need to be forgotten: It’s time to write down frothy valuations and move forward with fundamental-based pricingQuotable Moments“This is an epic monopoly like we’ve never seen. Think how much ChatGPT already dominates our lives. It’s the Standard Oil of tech.” — Jason discussing OpenAI’s market dominance“You’ve got this company with $4.5 trillion market cap with only six customers. That’s bad news. But the good news is all six of them are determined to spend themselves into oblivion to win the prize.” — Rory discussing NVIDIA’s customer concentration risk“Any hot AI deal, there is almost no diligence provided, nor is any done. It’s just done on Saturday. All you can lose is one extra money.” — Jason Lemkin on current venture funding practices“Having an early success is highly correlated with future success. Partly you get the referral effect, but partly I think it’s that you just have the stomach to roll the dice and you get braver.” — Rory on venture capital success patterns“I’m 0% cash, just like 2008. Nothing was more fun than selling my stock at a 70% loss to fix the roof on my house.” — Jason reflecting on his current market position“I’m seeing less diligence now than I did in 2021.” — Harry discussing the current state of venture due diligence“The hardest thing I find is honestly we are getting it wrong. A lot of our startups at 20VC Fund now are are growing nicely and I cannot predict what the next round wants because it seems to be moving so much and I’m having real problem predicting financing markets.” — Harry on the unpredictability of current funding marketsThanks for reading The Secrets To Scaling in The Age of AI! Subscribe for free to receive new posts and support my work. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

How to scale from startup to $300,000,000+ ARR by mastering the fundamentals of go-to-market strategyBuilding a billion-dollar B2B business isn’t about finding secret hacks or silver bullets. They don’t last or scale. It’s about mastering fundamental principles and being willing to reinvent yourself every six months to a year as you scale. Gaurav Agarwal, COO of ClickUp came to SaaStr Annual + AI Summit to share how they did it — and keep doing it.As someone responsible for “all things money” at ClickUp – sales, marketing, growth, pre-sales, and post-sales – Gaurav has lived through the reality that what gets you to $1M ARR is completely different from what gets you to $10M, $50M, $100M, $300M and beyond. Nothing scales infinitely, and every stage requires its own playbook.Here are the key principles that have driven ClickUp’s remarkable growth:1. Know Where You Win: The LTV vs. TAM MatrixMost companies fail because they try to adapt everyone else’s strategies without understanding their own fundamental positioning. Before you copy anyone’s playbook, you need to map your business on a simple 2×2 matrix:* X-axis: Customer Lifetime Value (LTV) – How much can you make from your customers?* Y-axis: Total Addressable Market (TAM) – How many customers are out there?This creates four distinct quadrants, each requiring completely different go-to-market strategies:High LTV, Small TAM: Whale Hunting You’re selling to Fortune 500 companies with limited prospects. Your channels must be high-touch: field marketing, trade shows, conferences, and business development. You can afford expensive customer acquisition because deal sizes justify the investment.Low LTV, Large TAM: Cast a Wide Net You can’t afford expensive acquisition channels. Focus on organic growth: content marketing, SEO, social strategies, and community building. You need LTV-to-CAC positive channels that scale efficiently.Low LTV, Small TAM: Exit Strategy If you have few customers who don’t pay much, you shouldn’t be in this business. Run away and find a better opportunity.High LTV, Large TAM: The Sweet Spot This is where ClickUp operates, and it’s the most exciting quadrant. You can make almost any channel work – enterprise sales teams, billboards, TV ads, digital marketing. The world is your oyster for distribution strategies.2. Learn From the Best-in-Class Across IndustriesDon’t limit yourself to studying companies in your vertical. The best growth strategies often come from unexpected sources:* For SEO: Study HubSpot, but also look at NerdWallet, Canva, and Zapier* For brand building: Don’t just look at B2B companies – examine what Liquid Death and Beats by Dr. Dre accomplished* For viral growth: Understand how consumer companies create shareabilityAt ClickUp, teams obsess over these best-in-class companies and adapt their learnings regardless of industry. A B2B company can absolutely learn from B2C growth tactics, because all customers are ultimately humans with similar psychological triggers.3. Avoid Zero-Sum ThinkingThe biggest mistake scaling companies make is creating false either/or narratives:* Product-led growth vs. sales-led growth – Why not both?* Brand building vs. demand generation – They should work together* B2B vs. B2C tactics – Customers are humans regardless of contextClickUp runs a dual-engine growth model that proves these approaches can be complementary:Product-Led Growth Engine: Focuses on users – acquire, activate, monetize, and expand. This gives ClickUp distribution at scale and catches a wide net of prospects.Sales-Led Growth Engine: Focuses on accounts where PLG has already landed users. The goal is reaching out to existing customers to drive expansion and deeper penetration.The results speak for themselves: when a customer moves from self-serve/PLG to sales-assisted, ClickUp sees an 11x lift in LTV. Product-led growth provides the distribution; sales-led growth maximizes the lifetime value.Most companies kill one motion in favor of the other, missing the massive opportunity of letting them feed into each other.4. Master Your Growth PortfolioAs you scale, you need to think like a portfolio manager. Your growth strategy should diversify across multiple bets – channels, segments, products, geographies – each with different risk/reward profiles.The 70-20-10 Resource Allocation Framework:* 70% on proven channels with high probability of success* 20% on small bets that deliver incremental gains* 10% on big bets like viral content or breakthrough strategies that could create inflection pointsThis approach delivers predictable growth with asymmetric upside. The 70% gives you financial predictability, while the 20% and 10% create opportunities for breakthrough growth moments.Key Portfolio Principles:Understand Risk vs. Reward: Higher risk channels might extend payback from 20 to 50 months, while optimizing for quick payback could limit top-line growth. Find the efficient frontier.Not All Channels Are Equally Repeatable: ClickUp can predict which SDR will hit their number and the ROI of Google branded search, but can’t predict which social media clip will get 50 million views. Plan accordingly.Channels Have Capacity Limits: After hiring 100 account executives, productivity per rep declines. Most paid channels hit diminishing returns, but some organic channels compound infinitely.5. Identify Compounding vs. Diminishing ChannelsUnderstanding which channels scale versus which hit walls is crucial for long-term success:Diminishing Return Channels:* Paid advertising (Google, Facebook, etc.)* Outbound sales beyond optimal team size* Most performance marketing channelsCompounding Channels:* Community: Takes time to build but becomes self-sustaining* SEO/Content: High domain authority compounds with each new piece* Viral Product Features: Like Slack or Dropbox, usage drives sharing* Network Effects: Value increases with each new userWhile you can’t scale Google paid search infinitely, you can scale Google organic content indefinitely with the right content engine.6. Never Assume You’ve “Made It”At $10M ARR, getting to $15M feels relatively easy. At $500M ARR, growing 50% to $750M is “extremely, extremely hard.” The only way to beat this challenge is maintaining hunger and assuming there are always better opportunities you haven’t discovered yet.Combat the Local Maximum Trap: Teams often get stuck optimizing their current approach instead of finding breakthrough strategies. Always assume better distribution tactics and motions exist.Apply Constant Pressure: Challenge yourself and your team to keep uncovering new ground. The moment you get comfortable is when competitors start gaining ground.7. Leverage AI for 10x Velocity GainsIf you view revenue generation as a complex machine, AI creates opportunities to optimize efficiency at multiple steps simultaneously. Companies not using AI to accelerate their go-to-market motion will be left behind.ClickUp’s AI-First Approach Across 80% of Revenue Functions:* AI SDR: Automatically reaches out to form submissions, generated $1M in pipeline in the first month* Content at Scale: Produces 40,000 pages monthly using AI, localized in 15 languages, resulting in 10x search impression volume* Customer Support: AI chat agents resolve issues faster, freeing human agents for proactive account managementThe key is systematically identifying every step in your revenue machine and asking: “How can AI increase throughput here?”The Top Scaling Mistakes to Avoid (According to Gaurav)Based on his experience scaling ClickUp to nine-figure ARR, Gaurav identified the most common and costly mistakes he sees companies make:1. Misunderstanding Your LTV/TAM Position and Copying Wrong StrategiesThe Mistake: Companies blindly copy successful strategies from other businesses without understanding whether they operate in the same LTV/TAM quadrant. A high-LTV, small-TAM business trying to use Canva’s mass-market SEO strategy will waste massive resources.The Fix: Map your business honestly on the LTV/TAM matrix first, then only study companies in your quadrant. A whale hunting business should learn from enterprise software companies, not consumer apps.2. Creating False Either/Or ChoicesThe Mistake: Forcing zero-sum decisions between complementary strategies. “We’re PLG, so we can’t do sales-led.” “We’re B2B, so we can’t use B2C tactics.” “We focus on demand gen, not brand building.”The Fix: Look for ways strategies can compound each other. ClickUp’s 11x LTV lift from combining PLG and sales-led growth proves these approaches can be synergistic, not competitive.3. Getting Stuck at Local Maximum and Stopping InnovationThe Mistake: Once companies find what works, they optimize that single approach to death instead of continuing to explore new opportunities. They become comfortable and stop pressure-testing for better distribution tactics.The Fix: Always assume there are better strategies you haven’t discovered. Maintain the 70-20-10 resource allocation where 30% of resources go toward new opportunities, even when current channels are performing well.4. Abandoning Compounding Channels Too EarlyThe Mistake: Companies give up on long-term compounding channels (community, SEO, viral features) because they don’t see immediate results, instead over-investing in diminishing return channels like paid ads because they’re more predictable.The Fix: Understand which channels have infinite scaling potential versus those with natural limits. Invest in building compounding channels even when they take longer to show results, while using diminishing return channels for predictable short-term growth.The Bottom Line: No Hacks, Just FundamentalsThere are no secret growth hacks for building a sustainable, ever-growing revenue machine. Success comes from:* Understanding your positioning in the LTV/TAM matrix and choosing appropriate strategies* Learning from the best-in-class regardless of industry* Avoiding zero-sum thinking and finding ways to combine approaches* Building a diversified growth portfolio with the right risk/reward balance* Investing in compounding channels while managing diminishing return channels* Maintaining hunger and constantly seeking better opportunities* Using AI to 10x velocity across every step of your revenue machineThe competition is getting fiercer every day. ClickUp is now accelerating at $300,000,000+ ARR. How about you?Thanks for reading The Secrets To Scaling in The Age of AI! Subscribe for free to receive new posts and support my work. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

The technical and product insights from Dara Ladjevardian’s AI cloning experiment at SaaStr Annual + AI Summit.The Clone Performance Reality CheckWhen Dara Ladjevardian, CEO of Delphi.ai, ran 1 million simulated conversations with digital versions of Brian Halligan (Chairman and founding CEO HubSpot), Lenny Rachitsky, Keith Rabois and Jason Lemkin, the most interesting findings weren’t about the business advice the clones gave—they were about how AI clones actually behave, fail, and succeed.And you can try them all yourself here:* Digital Jason, try it here* Digital Lenny, try it here* Digital Brian, try it here* Digital Keith, try it hereKey Technical Learnings1. Context Dimensions Drive Dramatically Different OutputsThe discovery: The same clone gives fundamentally different advice based on just four input variables:* Company stage (0-1M, 1-10M, 10-100M ARR)* Market context (emerging vs. established, crowded vs. uncrowded)* Team dynamics (solo vs. co-founder, data-driven vs. visionary)* AI adoption positionWhat this means technically: Current LLM approaches that treat context as simple “system prompts” miss the nuanced way human experts actually adjust their thinking. The clones needed sophisticated context weighting to perform authentically.The failure mode: Without proper context handling, AI clones default to generic advice that sounds like the person but lacks their actual decision-making sophistication.2. Temporal Knowledge Graphs Beat Static TrainingDara’s architecture insight: “The best way to represent a network of ideas that changes over time is a temporal knowledge graph.”Why this matters: A static knowledge graph might show Keith Rabois believed X in 2015, but his 2024 graph shows he believes Y. The temporal system tracks belief evolution to predict future responses.The technical challenge: Most AI clones train on a person’s entire corpus as if their views never changed. This creates internally inconsistent outputs that feel “off” to people who know the subject well.Real-world impact: Dara’s grandfather’s clone could apply 1970s Iranian business principles to 2024 AI startup decisions—something impossible with static training.3. Model-Agnostic Architecture Outperforms Single-Model TrainingStrategic decision: Delphi doesn’t train custom models—they use multiple existing models with sophisticated mind mapping on top.The reasoning: “We could train our own model right now, but why spend all that money? The product works really well by mapping out the mind and leveraging multiple models.”Performance insight: The hard problem isn’t the LLM—it’s accurately representing someone’s decision-making patterns and worldview. Once you solve that, you can ride the commodity curve of improving foundation models.4. Two-Mode Architecture Solves the Accuracy vs. Utility Trade-offStatic Mode: Only answers questions the person has explicitly answered before. Higher accuracy, limited utility.Adaptive Mode: Can predict responses to novel situations based on learned patterns. Higher utility, requires stronger guardrails.The professional insight: For doctors and lawyers, wrong answers create lawsuit risk. For creators and advisors, novel insights create value even if occasionally wrong.Product learning: Users need explicit control over this trade-off rather than a single “accuracy” dial.5. Identity Verification is Critical for Trust (And Scaling Pain)Current process: Every user submits photo holding ID. Dara manually verifies each one.The scaling problem: Manual verification obviously doesn’t scale, but automated systems miss edge cases that matter for trust.Why it matters: Creating clones of others without permission isn’t just unethical—it destroys platform credibility when discovered.The unsolved challenge: How to verify identity at scale while preventing abuse and maintaining trust.6. Guardrails Need Domain-Specific TuningThe discovery: Generic content filters don’t work for professional AI clones. A doctor clone needs different guardrails than a business advisor clone.Technical challenge: Building “pretty strict guardrails” requires understanding not just what the person would say, but what they’re legally/ethically allowed to say in their professional capacity.Performance impact: Over-aggressive guardrails make clones feel robotic. Under-aggressive ones create liability risks.What Actually Works in Clone ArchitectureThe Ray Kurzweil MethodBased on “How to Create a Mind” (2014): The brain is “a hierarchy of pattern recognizers.” Since LLMs are pattern recognizers, you can recreate minds by mapping patterns correctly.Focus on Representation, Not TrainingThe breakthrough insight: Spend engineering effort on accurately modeling someone’s thinking patterns, not on training custom language models.Multi-Model RedundancyUse multiple foundation models rather than relying on a single custom-trained model. The mind representation layer handles consistency.The Unsolved ProblemsBelief Evolution TrackingHow do you automatically detect when someone’s views have changed on a topic without manual annotation?Context Sensitivity at ScaleHow do you handle the exponential complexity of context combinations as you add more dimensions?Authenticity ValidationHow do you measure whether a clone “sounds like” the original person beyond basic accuracy metrics?Dynamic Guardrail AdjustmentHow do you automatically tune safety constraints based on the professional context and risk tolerance of different use cases?Additional Learnings from the Clones:* Speed vs. Accuracy Trade-off: The clones performed best when optimized for speed of response rather than perfect accuracy. Users preferred fast, “good enough” responses over slow, perfect ones.* Personality Quirks Matter More Than Content: The clones that felt most authentic weren’t the ones with the most training data, but those that captured subtle personality traits—Keith’s contrarianism, Jason’s directness, Brian’s optimism.* Context Switching is Expensive: When conversations jumped between topics (stage advice to pricing strategy to hiring), clone performance dropped significantly. Single-topic conversations maintained much higher quality.* The “Not Always Right” Disclaimer Actually Improves Trust: Dara noted that clones explicitly saying “I’m not always right” made users more comfortable engaging deeply, similar to how GitHub Copilot’s 35% acceptance rate doesn’t hurt adoption.* Users Test Clones Immediately: The first questions people ask are always attempts to “break” or test the clone with edge cases, not genuine advice-seeking. The clones needed to handle skeptical users before helpful ones.* Memory Across Conversations Doesn’t Scale: While individual conversations stayed coherent, maintaining context across multiple sessions created exponential complexity that hurt performance more than it helped.The biggest technical insight: Building authentic AI clones isn’t about better language models—it’s about better models of how humans actually think, decide, and evolve their beliefs over time.Thanks for reading The Secrets To Scaling in The Age of AI! Subscribe for free to receive new posts and support my work. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

We're back on 20VC + SasStr with Harry Stebbings, Jason Lemkin, Rory O'Driscoll, and special guest Jeff Lawson (Founder & Former CEO, Twilio)Bottom Line Up FrontRory O'Driscoll: Tesla's trillion-dollar pay package for Elon is a board betting everything on doubling down - they believe without him, the stock drops 75% overnight. It's intellectually coherent but terrifying risk concentration.Jason Lemkin: We're in the greatest wealth hunt in venture history. Orders of magnitude larger deals are the new normal. A $10 billion company feels "niche" today when we're discussing $100+ billion valuations.Jeff Lawson: The AI wave creates unprecedented opportunities for infrastructure companies like Twilio that aren't selling seats - no innovator's dilemma. SaaS companies selling seats face existential disruption as AI eliminates 75% of human roles.Harry Stebbings: Late-stage AI investing has become the rational play for VCs - when only valuation risk remains, even $100M checks into $13B rounds make mathematical sense for portfolio construction.The Trillion-Dollar Elon Bet: Rational or Reckless?Tesla's board just approved what could become the first trillion-dollar executive compensation package in history. But is this visionary leadership investment or a high-stakes gamble gone wrong?The Board's Logic: Double or NothingRory O'Driscoll dove deep into Tesla's 332-page proxy filing to decode the board's thinking. "Compensation is how boards reveal their real priorities," he explained. "Nothing else matters as much. The board wants the Elon bet - they believe they owe him the past and they're betting on him for the future."The package's operational metrics tell the story: Tesla needs to hit $400 billion in EBITDA (four times Google's current profitability), manufacture 20 million cars, deploy 10 million Full Self-Driving systems, and produce 1 million Optimus robots. It's essentially asking Elon to double the existing business while building entirely new categories.The Downside Protection TheoryJeff Lawson raised the critical counterpoint: "Maybe this isn't about upside - it's about the downside case. Tesla is overvalued as a car company. If valued purely on automotive fundamentals, it's worth 25% of current market cap. The other 75% is Elon's special sauce."This creates a prisoner's dilemma for the board. As Rory noted, "If you try to demonstrate resolve and he threatens to walk, you're down 75% next morning. The individual shareholders who voted for this compensation twice want to make this bet, even though it makes my head hurt."The New Benchmark for Founder CompensationJason Lemkin sees this setting a new standard: "This is the new normal for anyone whose board consists of their brother-in-law and other relatives. Almost all my portfolio companies - the founders control the board, not just from a cap table perspective, but from a relationship perspective."The trend is clear: instead of 2-3% top-ups after years of struggle, founders are now getting 7-8% packages tied to massive outcome targets of $10-100 billion valuations.Ramp Hits $1B ARR, Brex at $700M: AI Tide Lifting All Boats?Two major fintech announcements dominated the week: Ramp crossing $1 billion ARR and Brex hitting $700 million ARR with 50% growth. But are these isolated successes or signs of broader market strength?The AI Money Flow EffectJason Lemkin argues it's the latter: "The AI boom is filtering further down the stack. We're seeing it in Broadcom, Cisco - where our grandpa learned to be an engineer. If you're a B2B company seeing nothing from AI, you get an F. There's so much money flowing through this system."The fish food metaphor resonated: "It's floating down to where it's dark in the ocean now. It's embarrassing if you can't get any of it."Financial Services vs. Software ValuationsRory O'Driscoll provided clarity on how to value these hybrid companies: "They have the margin profile and core dynamics of financial services but the growth rate of software companies. Once growth slows, they'll be valued just like AmEx. The growth is what's saving them."Jeff Lawson offered the infrastructure perspective: "There's money out there, and it's got to go into some bank. Are they winning market share from legacy companies? Part of it. But during the mobile boom at Twilio, we had customers spending millions who didn't make it - that revenue went away and had to be replaced."Sierra's $10B Valuation: Bubble Territory or Brilliant Bet?Brett Taylor's Sierra commanding a $10 billion valuation at $100 million ARR (100x multiple) sparked intense debate about AI valuations and whether we're seeing rational investing or complete market detachment.The Generational Talent PremiumJason Lemkin made the bull case: "If Scale was for sale for $28 billion, Brett's got to be worth $56 billion. This is a generational guy who turned down being co-CEO of Salesforce to do this. You get everything - the ex-CTO of Salesforce and Facebook, his team, and a potential category leader."Rory O'Driscoll walked through the investment logic: "Is there a category that can support a big winner? Customer support is the number one AI use case outside coding. Are they going to be a winner? Clearly one of a small number with a great position. The only question left is: are you getting paid for the risk?"The Portfolio MathHarry Stebbings questioned the concentration: "Neil Mather is probably putting 10% of his fund into this next check. That's notable given the percent of his fund this represents."Jeff Lawson provided founder perspective: "When I start a company, that's 100% of my capital allocation - for my life, time, bank account, everything. Don't say you're brave putting 10% of the fund into one deal when the founder is putting 100% with no way out."Customer Support: The Perfect AI Use CaseThe consensus emerged around customer support being ideal for AI disruption. As Rory explained: "You have lots of people answering phone calls and emails. You can do 70-80% with AI. It saves money on a cost center. This is going to happen."Kleiner Perkins' $100M Anthropic Bet: Logo Deal or Smart Money?Kleiner Perkins investing $100 million into Anthropic's $13 billion round at a $183 billion valuation raised eyebrows as their first model provider investment.The "Logo Deal" DebateJason Lemkin initially called it a logo deal: "You can't walk into LP meetings without Anthropic or OpenAI on the website." But deeper analysis revealed more complexity.Rory O'Driscoll pushed back: "I don't think anyone does a $100 million logo deal. When you run the math, one of two things will happen: either the fastest slowdown in history in the next two years, or if this trajectory continues, this round will work."The Evolution of Venture CapitalThe discussion revealed how venture capital has fundamentally changed. Jason noted: "This isn't venture capital as we knew it. Venture capital today is 20% old school and 80% late-stage growth investing that would have been Fidelity."Jeff Lawson quoted Gloria Swanson: "I didn't leave Hollywood. Hollywood left me. This is venture capital today."Risk-Adjusted Returns in AIHarry Stebbings made the mathematical case: "When only valuation risk remains, you're putting $100 million in and potentially getting $300 million back. That feels like an easier way to make money than being in the trenches in the seed round."OpenAI's $10B Secondary: 1,000 New Millionaires Hit the MarketOpenAI's unprecedented $10 billion secondary sale creates the largest private liquidity event in history, potentially minting over 1,000 new millionaires in San Francisco.Historical Context and Market ImpactJeff Lawson recalled similar dynamics: "I remember feeling this way before Twitter's IPO. I was looking for our first house and thinking I had to buy before the IPO because everything's going to go nuts."Rory O'Driscoll provided perspective: "If this were a public company worth half a trillion dollars, 20% held by management means $100 billion. People selling 10% to diversify is entirely rational. It's only anomalous because it's private."The Recruiting Arms RaceJason Lemkin highlighted the competitive implications: "The impact on recruiting is so much bigger this generation. If you're running a boring B2B company doing triple-triple-double-double, even 36 months ago you were S-tier. Today, you're not going to get engineering talent when everyone's making eight figures."Jeff Lawson offered historical perspective: "It's like asking Jim Farley from Ford how he recruited developers during the teens when they could have worked at Twitter and Facebook. Maybe the key is getting out of Silicon Valley."Anthropic's $1.5B Author Settlement: Setting Legal PrecedentAnthropic's $1.5 billion payment to authors for copyright infringement established important legal boundaries for AI training data.The Legal FrameworkRory O'Driscoll analyzed the court ruling: "The judge said if you bought the book and paid $15, using it for training is legal. If you downloaded from pirate sites without paying, you get fined $3,000 per book. It's fairly coherent - 500,000 books at $3,000 each."The precedent is clear: AI companies must purchase content for training rather than downloading from pirate sites. "This wasn't cutting corners," Jason noted. "This was pure piracy - going to Pirate Bay for books."Future ImplicationsThe ruling suggests a path forward: legitimate content licensing for AI training. As Rory explained, "There will be a corpus available where someone has bought 500,000 books, scanned them, creating a legally compliant dataset for training."The Great Corporate AI Investment WaveFrom ASML's $14 billion investment in Mistral to Atlassian's $610 million acquisition of Browser Company, corporate America is making aggressive AI bets.ASML's Mysterious Mistral InvestmentThe semiconductor equipment giant's investment in the French AI company puzzled observers. Jeff Lawson explained the corporate cash dynamics: "When you're generating massive cash, it's orphaned on your balance sheet. You can't hire 1,000 engineers without hurting EPS. But swapping one asset for another can be essentially free if it doesn't decline."Rory O'Driscoll suggested European sovereignty concerns: "Non-US regions feel the need for local champions. It's the AI version of defense equipment - Europeans make weapons they have no business making from an economies of scale perspective because they don't want to rely on Americans."Atlassian's Browser Company AcquisitionThe $610 million cash acquisition sparked debate about corporate desperation versus strategic vision.Jeff Lawson was skeptical: "The thesis of needing a different browser for work didn't resonate with me. I'm not sure there are enough upsides to change behaviors."Jason Lemkin saw it as symptom of broader urgency: "Everyone feels like they have to make a play. Sometimes when you're itchy, you might not make the ideal investment."The SaaS Disruption ImperativeJeff Lawson delivered perhaps the most important insight of the discussion: "Any SaaS company today is primed for disruption because AI will decimate their seat base. AI will do the jobs people are sitting there doing in Atlassian products today."His advice was direct: "I would skate directly there and say: what is the job humans are doing in Atlassian products, and here's the AI version of that. That's what I would be doing."The Developer API EvolutionJeff Lawson shared his framework for successful developer companies, identifying only three categories that achieve breakaway revenue:Business Development as a Service"Developers can't open bank accounts or strike deals with AT&T on behalf of their company. But with Twilio, Stripe, AWS, you can engage in business relationships you weren't able to do previously."Capex as a Service"A developer can't spend $10 million to build a data center, but they can put it on a credit card. AWS and Google Cloud fit here."Algorithm as a Service"The algorithm must be so complicated that developers say 'I'm not smart enough to figure that out.' I used to only put DynamoDB here - infinite scaling databases are so hard. Now I'd put inference in that category, except it's becoming open source."The framework explains why many developer tools struggle: "Developers take your cool thing as a challenge. You're saying they can't build what you built. Especially when it gets to meaningful revenue - if you're paying $5 million a year, developers think they can save the company money by recreating it."The Fraud and Ethics DiscussionThe arrest of IRL's CEO for fraud prompted discussion about startup ethics in the current environment.The Enforcement PerspectiveJason Lemkin advocated for stronger consequences: "Venture rounds are getting done on Saturdays. Diligence isn't even being attempted. The best control would be if more founders who committed fraud went to jail."He continued: "There's no consequence to standing up at a top accelerator saying you have millions in revenue when it's not there. If a few more people went to jail, it would have the proper chilling effect."The Due Diligence ResponsibilityJeff Lawson offered a more nuanced view: "Whenever I read these fraud stories, I've never heard of the companies. If I've never heard of them as a real human being, maybe there wasn't much real behind them."Rory O'Driscoll balanced both perspectives: "The commission of crime is on the 22-year-old who lies. But the 40-year-old running money who's sophisticated owes the system a duty of care to check this stuff and not get carried away."Key TakeawaysOn Founder Compensation* Elon's package sets new benchmark for founder-controlled boards* Risk concentration reaches extreme levels with $1 trillion packages* Boards face prisoner's dilemma: pay up or risk 75% stock declineOn AI Market Dynamics* Infrastructure companies positioned better than SaaS (no innovator's dilemma)* Customer support emerges as perfect AI use case* Late-stage AI investing becomes mathematically rational for VCsOn Market Valuations* 100x ARR multiples justified for generational talent in large markets* Geographic sovereignty concerns drive irrational investments* Corporate cash seeking AI exposure creates acquisition opportunitiesOn Industry Evolution* Venture capital fundamentally changed: 80% late-stage growth investing* Developer API success requires business development, capex, or algorithm as service* SaaS companies face existential threat from AI eliminating seat-based revenueMost Quotable MomentsRory O'Driscoll* "Compensation is how boards reveal their real priorities. Nothing else matters as much."* "The board wants the Elon bet. If they try to demonstrate resolve and he threatens to walk, you're down 75% next morning."* "I didn't leave Hollywood. Hollywood left me. This is venture capital today."Jason Lemkin* "This is the greatest wealth hunt in venture history. A $10 billion company feels 'niche' today."* "If you're a B2B company seeing nothing from AI, you get an F. There's so much money flowing through this system."* "Venture rounds are getting done on Saturdays. Diligence isn't even being attempted."Jeff Lawson* "Don't say you're brave putting 10% of the fund into one deal when founders are putting 100% with no way out."* "AI will decimate SaaS seat bases. AI will do the jobs people are sitting there doing in Atlassian products today."* "You don't start companies to make money. You start companies because you think the world needs what you're building."Harry Stebbings* "When only valuation risk remains, valuation risk expands to fill like a vacuum."* "We're in a weird world where you can't buy the things that are great. You can't buy Sierra."* "If you're not in, you can't win. But oh my god, it's hard to know which bets to make." This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

From the SaaStr Annual / AI Summit – How three breakout AI companies rewrote the rules of enterprise sales. And see everyone at 2026 Annual + AI Summit May 12-14 2026 and SaaStr AI London Dec 2-3!Speaker BiosTalia Goldberg – Partner, Bessemer Venture PartnersTalia leads AI investments at Bessemer and has been at the forefront of understanding how AI companies break traditional SaaS metrics and business models.Kelly Loftus – Head of Startup Sales, AnthropicKelly has scaled Anthropic’s startup sales team from fewer than 10 people to over 150 as the company grew from 250 to 1,300 employees in just 18 months.Jacob Jackson – Machine Learning Engineer, Cursor (formerly OpenAI, Tab9, Super Maven)A veteran of the AI coding space, Jacob has been building developer tools since 2018 and joined Cursor 8 months ago after working as a researcher at OpenAI.Gorkem Yurtseven – CTO and Co-Founder, FAL (Features and Labels)Gorkem leads the technical vision at FAL, the generative media platform that hosts open and closed source image and video models via easy-to-use APIs.Top 5 GTM Takeaways* No Quotas, No Problem: Both Anthropic and FAL have completely abandoned traditional quota systems in favor of “shadow targets” due to unpredictable AI-driven growth patterns.* Technical Sales Teams Are Everything: All three companies prioritize hiring technically sophisticated sales teams that can use their own products and understand complex technical buyers.* Product-Led Growth Dominates: With massive inbound demand, these companies focus on fulfilling demand rather than generating it, requiring fundamentally different sales motions.* Shorter Planning Cycles Win: Traditional annual planning is dead—these companies are moving to quarterly or monthly targets due to rapid model improvements driving unpredictable adoption.* Internal AI Usage = Competitive Advantage: Companies eating their own dog food internally create better products and more credible sales conversations.The traditional B2B/SaaS sales playbook may not officially dead—but it is at least according to three of the hottest AI companies on the planet. In a revealing panel discussion, leaders from Anthropic, Cursor, and FAL pulled back the curtain on how they’ve built hypergrowth go-to-market engines without quotas, with technical sales teams, and powered by product-led growth that would make traditional SaaS executives’ heads spin.The Great Quota RebellionThe most shocking revelation came early: none of these companies use traditional sales quotas. Kelly Loftus from Anthropic dropped the bombshell first: “We still don’t really have quotas. We have shadow targets.”Why? “It’s really hard to predict exactly what is happening. The adoption is fast. A lot of this is driven by model intelligence, which you cannot predict over a long time period.”FAL’s experience was even more dramatic. “Beginning of this year, we were looking to hire a head of sales,” Gorkem shared. “Any good head of sales candidate was trying to negotiate a quota system. We thought doubling next year would be a good target. During the interviews and negotiations, we grew maybe 50%. We were almost halfway there already. We decided this is useless. We are not doing quotas.”Both companies are experimenting with shorter-term accountability—quarterly or monthly targets instead of annual quotas—because the pace of AI model improvements makes longer-term predictions meaningless.Technical Sales Teams: The New RequirementAll three companies have made a fundamental bet on technical sales teams—and it’s paying off massively.“We have a very technical sales team,” Jacob from Cursor explained, “partly because it’s a technical product, but also because there are a lot of ways the sales process can be accelerated with software and with Cursor.”This isn’t just about understanding the product—it’s about being able to use AI tools to accelerate the sales process itself. Cursor’s sales team actively uses their own product to build tools that help with sales, creating a virtuous cycle of internal usage and external credibility.Anthropic has scaled from fewer than 10 go-to-market people to over 150 as the company grew from 250 to 1,300 employees in just 18 months. Kelly’s approach focused on building for scale from day one: “When I joined, we did not have the concept of quotas. What I did was let’s just build a team around feedback, knowing this team is going to scale from 10 people to hundreds.”Product-Led Growth on SteroidsThe demand dynamics for AI products have created a fundamentally different go-to-market reality. Instead of generating demand, these companies are primarily focused on fulfilling it.“At Cursor, many of our first enterprise customers bought Cursor because their developers came to their management and they said we need this tool—or in many cases they were already using it,” Jacob revealed.This bottom-up adoption pattern means traditional enterprise sales motions are less relevant. “There’s so much demand for AI that people don’t need these massive sales teams and they can get things done with much leaner teams,” Gorkem observed.The result? Revenue per employee ratios that “totally break the brain” according to Talia from Bessemer, with all three companies achieving growth metrics that shatter traditional SaaS benchmarks.Internal AI Usage as Competitive AdvantagePerhaps the most overlooked aspect of these companies’ success is how aggressively they use AI internally—creating better products and more authentic sales conversations.Anthropic’s Knowledge Bot: “One of our favorite use cases is a Slack channel where employees can ask questions, and Claude searches our internal knowledge bases and answers,” Kelly shared. “It’s been extremely useful for productivity and time to onboard, especially across time zones.”Cursor’s Background Agents: Jacob highlighted their most advanced internal use case: “You can give tasks to AI that it will complete asynchronously. If it’s 90% right and 10% is off, you can easily drop into what the AI has been doing and correct it.”FAL’s Research-Driven Hiring: Gorkem revealed an innovative approach: “We hired maybe four people through our research grants program. You send us an email with a project, we give you compute for a couple of weeks with no strings attached. People have been doing great projects, and we ended up hiring them.”The New Metrics That Actually MatterWhen traditional SaaS metrics break down, what do you measure instead? Each company has evolved different North Star metrics:FAL focuses on customer concentration: “We care about big logos, but we want to make sure revenue is coming from at least 30 to 35 different companies rather than being concentrated at the top.”Cursor prioritizes product truth: “The thing we care about most is whether we personally want to use it in our day-to-day life. Revenue lags behind users, and users lag behind the fundamental quality of the product.”Anthropic emphasizes feedback loops: Rather than traditional sales metrics, they focus on “getting feedback on our models and continuing to work with partners to push model capabilities forward.”The Economics Behind the RevolutionThe fundamental reason these companies can abandon traditional sales playbooks comes down to unit economics that traditional SaaS executives would find terrifying—and liberating.“Before, if you were selling a SaaS product, the marginal cost was very little,” Gorkem explained. “But with AI, everyone has less margins because it comes with a real cost to serve each customer.”However, this trade-off enables something powerful: value-based pricing that scales with customer success. “When I first started selling developer tools for $49, I thought, ‘How much does this need to increase productivity to be worth it?’ It’s like 1% and it’s worth it,” Jacob noted. “Many people have been accelerated more than 2x already.”As Talia summarized: “COGs are the new CAC. You can spend a lot on cost of goods sold, but it means you can’t be spending a lot on customer acquisition because if you have low margins and really high acquisition costs, that’s tricky. But the good news is all of your products kind of sell themselves.”The Symbiotic Competition ModelPerhaps the most fascinating dynamic revealed was the relationship between Cursor and Anthropic—simultaneously customer/supplier and collaborative competitors.“We want to partner with companies like Cursor to drive the models forward,” Kelly explained. “Cursor has given us feedback on our models in the coding area and had access to our models before we released them.”Jacob reciprocated: “Whenever the models get better, we’re very happy because it means Cursor becomes more valuable to our users.”This collaborative competition represents a new paradigm for B2B relationships in the AI era, where ecosystem advancement benefits all players.Implications for B2B LeadersThe lessons from these three companies extend far beyond AI:* Question Annual Planning: In rapidly evolving markets, shorter planning cycles may be more effective than traditional annual quotas and targets.* Hire Technical Sales Talent: As products become more sophisticated, sales teams need deeper technical understanding to be effective.* Embrace Product-Led Growth: When possible, focus on fulfilling demand rather than generating it through traditional sales and marketing motions.* Use Your Own Product: Internal usage creates better products and more authentic customer conversations.* Collaborate with Competitors: In emerging markets, ecosystem advancement can benefit all players.The traditional SaaS sales playbook assumed predictable growth, high gross margins, and demand generation challenges. AI companies operate in a world of unpredictable hypergrowth, real marginal costs, and demand fulfillment opportunities.The companies that recognize this shift earliest—and adapt their go-to-market strategies accordingly—will define the next era of enterprise software.Quotable MomentsGorkem Yurtseven (FAL): “We thought doubling next year would be a good target for quotas. During the interviews, we grew maybe 50%. We were almost halfway there already. We decided this is useless. We are not doing quotas.”Kelly Loftus (Anthropic): “We still don’t really have quotas. We have shadow targets. It’s really hard to predict exactly what is happening when adoption is fast and driven by model intelligence you cannot predict over a long time period.”Jacob Jackson (Cursor): “Many of our first enterprise customers bought Cursor because their developers came to management and said we need this tool—or in many cases they were already using it.”Talia Goldberg (Bessemer): “COGs are the new CAC. You can spend a lot on cost of goods sold, but you can’t be spending a lot on customer acquisition when you have low margins and high acquisition costs.”Thanks for reading The Secrets To Scaling in The Age of AI! Subscribe for free to receive new posts and support my work. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

This week Canva Co-Founder and COO Cliff Obrecht joined Harry Stebbings, Rory O’Driscoll and Jason with honest, unfiltered insights on scaling to 240 million users, navigating AI transformation, and preparing for public markets.Canva’s CoFounder Joins 20VC + SaaStr on The Coming IPO, Where AI Works Today, Employee Liquidity, Figma’s IPO, and Much More!After 13 years building Canva into a $4 billion ARR juggernaut, Cliff Obrecht’s key insight is deceptively simple: “In the end, the thing that bails out our incompetence is your growth rate.” Whether facing 50x valuations in 2021 or 10x today, the fundamentals remain constant — compound growth covers a multitude of sins, while everything else is just noise.The $4B ARR Reality Check: What Actually Drives Growth at Scale — 90% is OrganicCanva will close 2025 “very close if not at $4 billion” in revenue, growing nearly 40% and reaccelerating after a post-2021 adjustment period. For Obrecht, this trajectory validates a contrarian approach to scaling that most B2B companies get wrong.“One thing you need to buck the trend of as you become a larger company is insular thinking and treating your user base like a wet tea towel that you need to ring out,” Obrecht explains. “90% of our user acquisition is organic.”The reacceleration story breaks down to three core drivers, with AI playing a surprisingly modest role:1. Core Flywheel Optimization (70%) “We just needed to reaccelerate all our core flywheels. We spoke about paying up for the team.”2. International Expansion (10%) “Going really heavy on international enhanced that.”3. AI Integration (20%) “I would probably say 20% [of reacceleration comes from AI].”This distribution challenges the narrative that AI is the primary growth driver for established SaaS companies. Instead, Obrecht’s thesis is that AI amplifies existing strengths rather than creating new ones.The AI Integration Playbook: Workhorses, Not GimmicksCanva runs billions of AI inferences monthly, but Obrecht’s approach differs markedly from AI-first companies. The philosophy: “We’re all about creating workhorses, not gimmicks. AI is just accelerating [our mission] massively, making it quicker, faster, and better for customers to achieve their goals.”The 10% GPU Tax Is RealLike Notion, Canva is paying the “GPU tax” — roughly 10% of revenue going to AI infrastructure and model providers. “100% yes, it already is [10% of revenue],” Obrecht confirms. “If you look at Lovable, their pass-through to Anthropic or model providers will be way more than 10%.”But this isn’t sustainable at current levels. Canva’s optimization strategy reveals how smart SaaS companies should think about AI costs:Near-term Reality:* Heavy compute costs for new AI products* Pass-through pricing to OpenAI, Anthropic, others* Unified credit model for usage managementLong-term Optimization:* “We’re betting on distilling these models down, understanding user queries and where I need the frontier model versus where I can deploy the model that’s on-device or self-hosted”* “Companies will get better at picking the right model for the right job”* “We view some of those upfront costs as more of a marketing cost than a long-term enduring cost of goods”The Credit Model SolutionFacing 240 million users and October’s “whole slew of new AI products,” Canva solved the margin problem with usage-based pricing layered onto subscriptions:* Free users: Limited AI credits* Premium subscribers: Expanded credit allocation* Heavy users: Usage-based pricing beyond thresholds“We need to maintain that margin,” Obrecht notes. This hybrid approach lets Canva capture AI value without destroying unit economics.The Billion-Dollar Balance Sheet StrategyPerhaps Canva’s most contrarian move: maintaining a billion-dollar cash position while profitable. “We’ve had a billion sitting on our balance sheet for ages — that’s a flex, people,” Obrecht says with characteristic directness.This wasn’t accident but strategy, informed by hard experience:The 2021-2022 Lesson:* 2021: $40 billion valuation at 50x revenue* 2022: Crashed to $26 billion as markets corrected* Lesson: “The markets will do what they’re going to do, but as a company, we can compound growth, compound margins, and deliver value to customers”The Insurance Policy Approach: “The truth is it’s a rocky journey. There’s probably some bumps ahead. I think most founders will be happier with a bigger balance sheet. The really great ones are the guys who can take the capital and then have the discipline not to use it foolishly.”For Canva, cash isn’t just defensive — it’s offensive optionality for when markets inevitably shake out competitors who “pissed it all away in performance marketing.”IPO Strategy: Employee Liquidity Trumps EverythingAfter 13 years and eight years of profitability, Canva doesn’t need public market capital. But employees need liquidity, and secondary markets aren’t cutting it.“We’re 13 years old as a company. Our employees should have liquidity. They’ve created all this value. How can we make it easy for them to access that wealth that’s built up?”The secondary market reality check is brutal:Problems with Private Liquidity:* “While secondaries, annual secondaries are a mechanism for that, it’s pretty janky”* “Particularly in some jurisdictions, it’s downright impossible”* Employees need employer permission for their own wealth access* Supply-demand imbalances when major companies (like Figma) go publicPublic Market Arbitrage: “The public markets now are valuing companies a lot higher [than private markets]. Ultimately the volume of capital dictates that multiple and there’s such an immense amount of capital being deployed in public markets that just is driving up those valuations.”Canva brought in Zoom’s former IPO CFO Kelly and is “gearing up to be an IPO ready company.” The timing question remains open, but the infrastructure is being built.Distribution Moats in the AI EraWhile AI companies fight for attention, Canva’s distribution advantages compound. The company is now the #5 most-referenced domain on ChatGPT and the #1 productivity app in the platform.“Over 5% [of images uploaded to Canva] now come from ChatGPT,” up from 0.02% eighteen months ago. This represents “LLM SEO” — winning discovery in AI platforms the same way companies optimized for Google search.But Obrecht’s insight goes deeper: “As soon as these LLMs started taking off, we had the conversation: is our SEO team working on LLM optimization? And there’s definitely a team at Canva working on that.”The lesson: Distribution advantages require active maintenance and adaptation, not passive hope.The Mercenary Problem vs. Mission AlignmentDiscussing Meta’s AI talent acquisition spree, Obrecht revealed Canva’s cultural philosophy: “When I was a B2B founder trying to be driven but touchy-feely, I was sort of anti-mercenary. If you’re not on my journey, I don’t want you. This is a long path — Canva’s been doing this for 20 years.”This mission-first approach creates different talent dynamics:Advantages:* Lower churn during difficult periods* Aligned decision-making during market volatility* Sustainable culture at scaleTrade-offs:* Potentially slower talent acquisition* Higher bar for cultural fit* May miss short-term optimization opportunitiesFor Canva, the trade-off makes sense: “Sometimes you need mercenaries and sometimes there’s a tool for the job. But I think this is… Zuck knows this is a bunch of mercenaries. Some of them are going to fall in battle.”Pricing Strategy: From Seats to ConsumptionAI is forcing SaaS pricing model evolution. Canva’s hybrid approach previews the future:“We need to rethink our seat-based pricing model because some of the tools, particularly around marketing tools we’re creating, enable a single marketer to deploy tens of thousands of pieces of content.”The new reality:* Traditional seats: Base functionality at fixed price* AI consumption: Usage-based pricing above thresholds* Value alignment: Pricing scales with customer outcomes“One person can do inordinate amounts of work and that’s using a huge amount of compute. You can’t charge 20 bucks a seat for that level of breadth.”Key Takeaways for B2B Leaders1. Growth Rate Forgives All Sins Market valuations will fluctuate wildly, but sustained growth compounds through any cycle. Focus on fundamentals, not noise.2. AI as Accelerant, Not Strategy AI should amplify existing strengths, not become the entire value proposition. Workhorses beat gimmicks at scale.3. Cash Is Strategic Flexibility In uncertain markets, balance sheet strength enables opportunistic moves while competitors struggle.4. Employee Liquidity Drives IPO Timing For profitable companies, public markets provide liquidity solutions more than capital access.5. Distribution Moats Require Active Defense Winning new platforms (LLMs) requires the same systematic approach as traditional SEO.6. Mission Beats Mercenaries at Scale Cultural alignment becomes more valuable as companies mature and face inevitable challenges.The meta-lesson from Canva’s journey: building at scale requires different muscles than early-stage growth. The companies that understand this distinction — and build accordingly — separate themselves from the pack.Cliff Obrecht co-founded Canva in 2012 and serves as Chief Operating Officer. Under his leadership, Canva has grown to 240 million monthly active users and approaching $4 billion in annual recurring revenue.The full convo here:Thanks for reading The Secrets To Scaling in The Age of AI! Subscribe for free to receive new posts and support my work. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

We had a great one this week — Marc Benioff joined Harry, Rory, and Jason on 20VC+SaaStr this week to deliver one of the most grounded and passionate takes on AI we’ve heard from any enterprise leader. In a market fueled by AGI promises and $10 billion funding rounds, Salesforce CEO’s cut through the hype while revealing his company is quietly building a billion-dollar AI business — by focusing on practical applications over futuristic fantasies.And Marc shared for the first time how AI is letting them follow up on 1,000,000+ leads their human sales team … never followed up on.Bottom Line Up FrontAI is working at enterprise scale, but not always in the way the hype machine suggests. Benioff’s Salesforce has achieved over $1 billion in AI and data cloud revenue—their fastest-growing product ever—by deploying agentic systems that actually solve customer problems today. Meanwhile, the venture ecosystem continues pouring unprecedented capital into AI infrastructure plays that may struggle to justify their valuations without dramatic changes in enterprise spending patterns.The Bottom Line Up Front:* Benioff’s Reality: “I don’t think that there will be a piece of software that we sell that will not be agentic.” Salesforce achieved $1B+ AI revenue faster than any product in their history by focusing on practical applications rather than AGI promises, while redeploying 4,000 support agents to higher-value roles.* Harry’s Concern: “I don’t feel like we’ve ever had the concentration of value tied to AI in seven companies as we have today.” The MAG-7’s unprecedented market concentration around AI creates systemic risk, while traditional growth metrics become meaningless when 10% growth at $40B scale adds an entire Palantir annually.* Rory’s Math: “You actually need these things to take vast chunks out of the labor budget and be worth 20, 30, $40,000 almost a head to the enterprise for the math to work.” Foundation model valuations require AI agents to capture massive enterprise labor budgets—a scale that current use cases haven’t yet reached.* Jason’s Evolution: “The fundamental architecture of an enterprise software company in the future is not exactly as it was in the past.” Companies must redesign organizational structures around AI capabilities, with 80% of VCs now refusing meetings with non-AI founders regardless of fundamentals.The Reality Check We NeededBenioff opened with a direct challenge to the AGI narrative: “You’re talking to somebody who is extremely suspect of anybody who uses those initials, AGI. I think that we have all been sold a lot of hypnosis around what’s about to happen with AI.”This isn’t technological pessimism—it’s operational realism. Benioff acknowledged AI’s power while stripping away the mysticism: “Large language models are two things. They are a finite set of algorithms… and a relatively finite set of data that has come off the internet. Those two things together really provide kind of the state of the art of large language models today.”The warning about over-reliance resonated particularly strongly. Benioff cited articles about doctors becoming “intellectually lazy” due to over-dependence on inaccurate AI, calling it “a huge warning sign for all of us around AI.”The $1 Billion Proof Point and the 100 Million Lead RevolutionWhile others chase AGI dreams, Salesforce is monetizing AI reality. Their data cloud and AI combination has exceeded $1 billion in revenue and represents their “fastest growing cloud product ever in 26 years.” This isn’t incremental feature revenue—it’s a fundamental platform shift.The numbers tell the story of practical AI deployment:* Reduced support agents from 9,000 to 5,000 through agentic systems* 100+ million historical leads now being contacted through AI-powered sales agents* AgentForce handling as many customer interactions as human support agents“This is a product that a year ago we hadn’t even announced. This is a product that wasn’t even shipped until November of last year,” Benioff noted, highlighting the unprecedented speed of enterprise AI adoption when the use cases actually work.The 100 Million Lead Follow-Up ChallengeThe most striking example of AI’s practical impact came from Benioff’s revelation about Salesforce’s own massive lead management problem. “Over the last 26 years, Salesforce has had more than 100 million people contact us that we’ve not been able to call back. We just have not had the people. That’s just all there is to it.”This wasn’t a technology problem—it was a human capacity constraint that plagued even one of the world’s most successful software companies. Despite having “like 15,000 sales people,” Salesforce simply didn’t have enough SDRs to handle the volume of inbound interest. The math was brutal: 100 million leads over 26 years represents nearly 4 million leads annually that went completely uncontacted.Think about the revenue implications. If even 1% of those 100 million leads could have converted to customers at Salesforce’s average deal size, that’s millions in lost revenue annually. This represents the classic enterprise software scaling problem: demand exceeding human capacity to respond.The AI SDR Solution in ActionBenioff described how their AI SDR system now tackles this previously impossible challenge: “We have this agentic sales now. And not only are we doing support, but this agentic sales is calling everyone back and having conversations with them and then deeply integrating it through the omni-channel supervisor into our new agentic sales product.”The system works through what Benioff called an “omni-channel supervisor”—an AI orchestration layer that coordinates between human agents and digital agents. This isn’t simple automation; it’s intelligent triage and routing that determines when human intervention is required and when AI can handle interactions independently.The AI SDR process includes:* Automated outreach to previously uncontacted leads from the 100 million backlog* Conversation management through natural language interactions* Qualification and scoring based on response patterns and engagement* Seamless handoff to human SDRs when conversations reach complexity thresholds* Integration with existing sales processes and CRM dataThe “Customer Zero” Validation StrategyBenioff emphasized that Salesforce serves as “customer zero” for their AI products, using their own 100 million lead challenge as the proving ground. This approach provides compelling advantages:* Credible case studies: When selling AgentForce, Salesforce can point to their own massive lead follow-up success* Real-world optimization: Internal usage reveals gaps and improvement opportunities before customer deployment* Sales confidence: Representatives can speak from direct experience rather than theoretical benefits* Risk validation: Salesforce absorbs the implementation risk before asking customers to do the sameThe Talent War RealityWhile Meta spends billions acquiring AI talent and Zuckerberg pays unprecedented premiums for teams, Benioff took a different approach: “No, and we’re not.” Instead of buying talent, Salesforce focused on redeploying existing headcount more effectively through AI augmentation.This represents a fundamental shift in enterprise architecture. As both Marc and Jason noted: “The fundamental architecture of an enterprise software company in the future is not exactly as it was in the past.” Rather than replacing workers wholesale, successful AI implementation enables workforce optimization and value chain elevation. Thousands of headcount can be redeployed to new functions.Palantir’s Enviable Pricing, and The Rise of the Forward Deployed EngineerThe Palantir discussion revealed perhaps the most telling moment of competitive envy from Benioff. Despite Salesforce’s $41 billion revenue dwarfing Palantir’s $4 billion, the market cap comparison stung: “That caught my attention. I’m like, how do I get that 100 times revenue multiple?”The pricing revelation was even more striking. Benioff’s admission that Palantir’s publicly available price list made him question his own pricing strategy speaks to a fundamental shift in enterprise software economics. “I’m actually delivering, like, I’m automating the whole VA at this price. Like, what would they be charging? I mean, my prices are low compared to theirs.”This isn’t just about individual deal sizes—it’s about market positioning. Palantir has successfully positioned itself as the premium AI solution for large enterprises, commanding prices that would be unthinkable for traditional SaaS products. They’ve broken through the conventional enterprise software pricing ceiling by selling transformation rather than tools.The Forward Deployed Engineer RevolutionThe forward-deployed engineer (FDE) concept represents a fundamental reimagining of the enterprise sales process. Traditional enterprise software follows a predictable pattern: demo, proof of concept, negotiation, implementation. Palantir flips this by deploying engineers before the deal is signed.“We don’t have that kind of branding of, these are our forward deployed engineers where now we’re going to start building your product now before we’ve really signed a deal,” Benioff observed. “And I think that idea is very cool that all of a sudden you’re like in there kind of saying, yeah, we’re going to make a bet that we’re going to start doing business together. So we’re going to start building now.”This approach solves the classic enterprise software chicken-and-egg problem. Customers can’t envision the solution until they see it working with their data, but vendors can’t build custom solutions without revenue certainty. FDEs bridge this gap by making the upfront investment in customer-specific development.The model works particularly well for AI implementations because:* Complexity Requires Customization: Unlike traditional SaaS where one-size-fits-most works, AI applications need deep integration with existing workflows and data structures.* Proof Points Drive Adoption: Executive buyers need to see AI working with their actual data and use cases, not generic demos.* Speed to Value: FDEs can achieve working prototypes in weeks rather than the months typically required for traditional enterprise software implementations.* Premium Pricing Justification: The white-glove service model justifies significantly higher price points than self-service or standard implementation approaches.Benioff’s interest in adopting the FDE model signals recognition that the enterprise software playbook is evolving. The companies winning the largest AI deals aren’t just selling software—they’re selling guaranteed outcomes through intensive human-AI collaboration during the sales process itself.The broader implication is that enterprise software companies may need to fundamentally restructure their go-to-market organizations. Instead of separate sales, engineering, and professional services teams, the future may belong to integrated teams that can sell, build, and implement simultaneously. This requires significantly different hiring profiles, organizational structures, and economic models—but potentially unlocks the premium pricing that has made Palantir one of the most valuable enterprise software companies despite its relatively modest revenue scale.The B2B Apps Survival DebateBenioff delivered his most passionate response when addressing suggestions that SaaS applications would become “just CRUD databases.” He called this “one of the greatest disservices that has been done to our whole industry” and “crazy talk.”His vision preserves application interfaces while adding agentic layers: “I need apps and I need agents and I need them to work together.” This represents the realistic middle ground between “AI replaces everything” and “nothing changes”—applications evolve with AI augmentation rather than disappearing entirely.The Anthropic $10B Round AnalysisThe discussion of Anthropic’s massive funding round revealed the disconnect between foundation model hype and enterprise adoption reality. While the round was “4x oversubscribed,” the math on market size remains challenging.Rory’s analysis was particularly illuminating: For foundation models to justify these valuations, “you actually need these things to take vast chunks out of the labor budget and be worth 20, 30, $40,000 almost a head to the enterprise for the math to work.”Even using Salesforce’s AI SDR as an example—potentially generating $3.6 billion in additional revenue with a 30% uplift—the LLM cost component at 20% of revenue would only represent $720 million flowing to foundation model providers. Scaling this across the enterprise software market suggests foundation model TAM may be smaller than current valuations imply.The Consensus Investment DebateMartin Casado’s tweet about consensus investing being “dangerous in early stage” sparked valuable discussion about AI investment strategies. The consensus view: most AI investments today are technically sound directional bets, even if individual companies fail.As Jason observed: “When you’re on this mega trend of an architectural replatforming, a goodly amount of the correct investments to do are fairly consensus in terms of the broad macro themes.”However, non-consensus bets face follow-on capital challenges. Jason’s advice to non-AI founders: “Don’t expect any money. 80% of the folks I can refer you to are not going to take your meeting.”Market Timing ConcernsThe concentration of value in AI-focused public companies raised reversion-to-mean concerns. As Harry noted: “I don’t feel like we’ve ever had the concentration of value tied to AI in seven companies as we have today.”Yet recent earnings from MongoDB, Okta, Box and others showing AI-driven reacceleration suggest the enterprise software incumbents may finally be capturing AI value. This could provide more sustainable growth than pure-play AI companies facing eventual margin pressure.Key Takeaways* Enterprise AI is real and scaling fast: Salesforce achieved $1B+ in AI revenue faster than any product in their 26-year history* Practical applications beat futuristic promises: Agentic support and sales systems deliver immediate ROI while AGI remains distant* Workforce evolution is here and accelerating: Successful AI deployment rebalances and elevates human capabilities rather than eliminating jobs. But headcount may be radically repurposed in many segments (support, SMB sales, SDRs, etc).* B2B apps will survive with AI augmentation: Applications evolve with agentic layers rather than being replaced by chat interfaces* Foundation model valuations face math challenges: Current market caps require enterprise AI spending at unprecedented scales* Consensus AI investing makes sense: Technical direction is clear even if individual companies face execution risk* Follow-on capital flows to AI: Non-AI companies struggle to raise subsequent rounds regardless of fundamentalsQuotable MomentsBenioff on AGI hype: “You’re talking to somebody who is extremely suspect of anybody who uses those initials, AGI. I think that we have all been sold a lot of hypnosis around what’s about to happen with AI.”Benioff on the future of software: “I don’t think that there will be a piece of software that we sell that will not be agentic.”Harry on market concentration: “I don’t feel like we’ve ever had the concentration of value tied to AI in seven companies as we have today. And I am looking at it now going like, I really hope there’s not a blip here.”Harry on growth at scale: “When you’re doing $40 billion, 10% growth is adding $4 billion, which is an entire Palantir every year.”Rory on venture dynamics: “Venture is a game played by 6,000 people, and in the end, Sequoia wins.”Rory on foundation model math: “You actually need these things to take vast chunks out of the labor budget and be worth 20, 30, $40,000 almost a head to the enterprise for the math to work.”Jason on workforce evolution: “The fundamental architecture of an enterprise software company in the future is not exactly as it was in the past, that the fundamental architecture of the company will be different.”Jason on follow-on capital reality: “80% of the folks I can refer you to are not going to take your meeting” (to non-AI founders).* This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

When you think about Product-Led Growth (PLG) success stories, few companies exemplify the model better than Calendly. Founded in 2013 by Tope Awotona, Calendly has grown from a simple scheduling tool to a scheduling powerhouse that's touched "double digits" of the world's billion knowledge workers - meaning over 100 million people have used the platform at some point.At SaaStr Annual + AI Summit, Awotona shared the hard-won lessons from building one of the most successful PLG companies of the last decade. What makes his insights particularly valuable is his transparency about the mistakes Calendly made along the way - and how they course-corrected.Today, Calendly maintains a fascinating 90/10 revenue split between self-serve and sales-led motions, with customers ranging from individual contributors to million-dollar enterprise accounts. But getting that balance right took years of experimentation, data analysis, and some painful lessons about when PLG and enterprise motions can cannibalize each other.Top 5 Key Learnings from Calendly's Journey1. The Free Plan Is Your Marketing Engine - Protect It at All CostsPerhaps the most counterintuitive insight from Awotona: Calendly spends "almost zero dollars on marketing campaigns." Their growth is entirely driven by user activity on the platform. Free users aren't just prospects - they're active marketing assets with measurable LTV."We're happy for people to use the product for free," Awotona explained. "Even free users have an LTV associated with them because we spend almost zero dollars on marketing campaigns."The pressure to tighten free plans is constant. Sales teams consistently identify the free version as their #1 competitor, not any external product. But since putting up their first paywall in 2014, Calendly has never removed features from the free plan - they've only made it more generous.The Takeaway: If your free plan drives viral growth, resist the short-term temptation to squeeze it. Instead, add more value to paid tiers rather than removing value from free ones.2. Viral Loops Get Harder at Scale - But Scale CompensatesCalendly closely tracks two critical metrics: meetings-to-signups conversion rates and signups-to-activation rates (defined as five people scheduling with a new user). As expected, these conversion rates decline at scale - the viral coefficient naturally decreases as you reach market saturation.But here's the key insight: "The good thing is the top line - the denominator - is getting bigger. That compensates a little bit for that degradation in conversion rate."The Takeaway: Don't panic when viral coefficients decline at scale. Focus on optimizing the absolute numbers while understanding that percentage-based metrics will naturally compress as you approach market saturation.3. Hybrid PLG + Enterprise Is Incredibly Hard to Get RightThis might be Awotona's most valuable insight for SaaS leaders. Calendly made both classic mistakes in balancing PLG and enterprise motions:Mistake #1: Under-investing in enterprise and leaving seven-figure deals on the table Mistake #2: Over-investing in enterprise and having the sales team cannibalize PLG revenue"What we found was the growth in our customer acquisition cost outpaced the incremental revenue growth. The enterprise business was really cannibalizing the PLG business," Awotona shared. By relaxing qualification rules to feed the sales team, they were simply converting self-serve prospects into sales-assisted deals - same revenue, higher cost, longer sales cycles.The Takeaway: Be "incredibly analytical" with holdout groups and rigorous testing. The superficial metrics might look good while you're actually damaging your more efficient channel.4. External-Facing Roles Are the Wedge Into EnterpriseCalendly's ideal customer profile focuses on external-facing roles: sales, customer success, recruiting. This represents about 25% of headcount in most organizations globally. But here's the strategic insight: these users become the wedge for enterprise expansion.Their largest customer - a financial services company doing $1M+ in annual revenue - started with "a few people in the company using it. They loved it and then decided to expand." The expansion took 6-8 months from a sub-$20K footprint to seven figures.The Takeaway: Identify the roles that get the most value from your product and use them as your enterprise wedge. Let the product prove itself with end users before enterprise sales gets involved.5. Product Development Resource Allocation Must Match Business Model EvolutionWith a 90/10 revenue split (PLG/Enterprise), how do you allocate product and engineering resources? Awotona's answer has evolved significantly:* Initially: 90/10 allocation matching revenue* 2020 Growth-at-all-costs era: Asymmetric allocation heavily favoring enterprise* Today: "The pendulum has swung back" based on product maturity and diminishing returns analysis"We look at the maturity of the product for different cohorts and analyze where we're hitting diminishing returns," he explained.The Takeaway: Resource allocation should be dynamic, based on product maturity and ROI analysis, not just revenue percentages. Sometimes your PLG motion needs more investment despite generating most revenue efficiently.Tope's Top 4 Mistakes (In His Own Words)1. Over-Staffing Enterprise Sales During Growth-at-All-Costs Era"We expanded that sales team greatly in 2020. Revenue grew, accounts over $100K grew, but customer acquisition cost outpaced incremental revenue growth. The enterprise business was cannibalizing the PLG business."2. Under-Analyzing the Hybrid Motion"I thought we were very analytical ourselves, but there's a lot more analysis we could have done, a lot more testing with holdout groups. When we ended up doing them, it was pretty eye-popping what we saw."3. Relaxing Sales Qualification Rules Too Much"We relaxed the qualification rules for letting people talk to a salesperson. What we would see is that revenue would convert, but if we just let them self-serve, a lot of them would have converted anyway - same revenue, but you add time and friction to the deal."4. Not Being Strategic Enough About Product Allocation"We've tried organizing teams by PLG vs Enterprise, by customer size, by capabilities. We've tried a number of different approaches. The most permanent design we've had is organizing by capabilities and having them own problems from individuals through enterprise."The common thread in all these mistakes? Not being analytical enough. Awotona's biggest recommendation for founders building hybrid PLG-Enterprise motions: "Be incredibly analytical, do a lot of testing, use holdout groups."What looks like success on the surface might be masking fundamental problems with your go-to-market efficiency. Sometimes the "best" sales reps are actually the worst when you analyze cohort retention. Sometimes growing enterprise revenue is actually just expensive PLG revenue in disguise.The companies that crack the hybrid code - like Calendly eventually did - are the ones willing to dig deep into the data and make hard decisions based on what they find, not what they hope to see.Thanks for reading The Secrets To Scaling in The Age of AI! Subscribe for free to receive new posts and support my work. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

Gusto co-founders Josh Reeves (CEO) and Eddie Kim (CTO) came to SaaStr AI Summit to share their journey building “Gus,” an AI assistant now used by hundreds of thousands of small businesses.Rather than chasing AI trends, they focused on solving real compliance pain points for small business owners navigating complex regulatory requirements across 50+ different state and local jurisdictions.Their approach involved two key tracks: conversational interfaces that make software more intuitive, and automation that eliminates time-consuming tasks entirely. The result is an AI system that generates reports, executes actions like approving time-off requests, navigates complex compliance requirements, and creates optimal shift schedules. Gusto’s “startup within a startup” methodology, 90-day roadmap horizons, and hybrid interface philosophy offer practical lessons for any SaaS company serious about AI implementation.Top 3 Takeaways from Gusto Co-Founders Josh Reeves & Eddie Kim:* Create a “Startup Within a Startup” for AI Projects – Gusto built Gus by creating an independent team that operated outside normal engineering processes, shipped weekly, and didn’t even use project tracking tools initially. This allowed for the rapid experimentation needed in the fast-moving AI landscape.* Focus on Problems Only You Can Solve – Be disciplined about which AI problems to tackle versus which ones the broader AI ecosystem will solve naturally. Gusto focused on Gusto-specific challenges while betting that general AI capabilities would improve on their own.* The Future is Hybrid Interfaces, Not Just Conversational – While conversational AI is powerful for many tasks, the best user experience combines conversational and graphical interfaces at the right moments. Not everything is better done conversationally.The Mission: Making Small Business Compliance Suck LessBefore diving into the technical details, Reeves made something crystal clear: “Companies don’t exist for the sake of it. We exist to go fix something, to go serve our customer and solve a pain point in their life.” For Gusto, that pain point is compliance hell.Small business owners navigate a maze of local, state, and federal regulations. “The US in particular is more like 50 countries than one when it comes to all the different rules and requirements a business owner has to navigate,” Reeves explained. This isn’t about using AI because it’s trendy – it’s about using AI to solve a very real, very expensive problem.Two Tracks: Conversational Interface + AutomationGusto approached their AI strategy along two clear tracks that every SaaS company should consider:Track 1: The Conversational Interface Revolution“We are navigating a pretty massive paradigm shift in how software gets used,” Reeves noted. The conversational interface paradigm shift means customers can simply talk to software instead of figuring out which buttons to click or pages to navigate.But here’s the key insight: this doesn’t replace traditional interfaces entirely. “We’re investing a lot of time in the Gusto web app, in our mobile app,” Reeves clarified. “We think they work in conjunction with each other.”Track 2: Pure AutomationThe second track focuses on literally automating tasks – both for internal team members (“Gusties”) and customers. “Think about something you spend 5, 10 minutes on. If it could just be done automatically for you, time savings is cost savings for a business owner.”What Gus Actually Does (Beyond Just Answering Questions)Kim broke down Gus’s capabilities, and they go far beyond typical chatbot functionality:Advanced Reporting & Analysis: Gus generates reports and provides critical business insights, not just raw data dumps.Action Execution: Tell Gus “I’d like to approve Sally’s time off request” and it will look up the request, show you details, and execute the approval upon confirmation.Compliance Navigation: With tens of thousands of compliance rules affecting small businesses, Gus tells you exactly which ones apply to your specific business based on Gusto’s deep knowledge of your company.Complex Scheduling: Gus can set up ideal shift schedules that automatically comply with break requirements, maximum daily hours, weekly limits, and other labor regulations.The result? Gus is “generally available to every single one of our customers on Gusto’s platform” as of last week, actively being used by hundreds of thousands of small businesses.The “Startup Within a Startup” PlaybookHere’s where Kim shared the tactical gold. Building AI products requires a fundamentally different approach than traditional SaaS development:Start Fresh: “We started with a brand new codebase from scratch” and “politely excused ourselves from many of our engineering team’s rituals and processes.”Move Fast, Track Later: “It wasn’t even until recently that we actually started to track our work in things like Jira and Asana.” Instead, they’d sit together, decide on priorities, and ship by week’s end.Embrace Uncertainty: “We would just sort of sit all together in a room, talk about what are the most important things for us to work on. And by the end of the week, we would ship it.”Three Hard-Won Lessons from Building GusLesson 1: Vision Clear, Roadmap Fluid“Although our vision for what problems we want to solve for the next few years is very, very clear… the roadmap of what we’re building cannot be more than 90 days out because we’re learning so much right now.”Every three months, they learn something completely new that changes what they need to build. Traditional annual roadmapping doesn’t work in AI.Lesson 2: Watch What Customers DON’T DoWhile traditional SaaS metrics matter, Kim emphasized something more valuable: “It’s even more important to observe what customers are not doing on your product… what they’re doing right before they actually start using your product.”He calls this “the work before the work” – the preparation customers do before using your product. “That is a gold mine of problems that you can be solving for your customers.”Lesson 3: No Single Interface Rules Them All“There is no single interface to rule them all. With the advent of AI, there’s so many things that are done better conversationally now, but not everything is done better conversationally.”The future is hybrid – conversational at the right moment for the right task, graphical when that makes more sense.The Internal Productivity MultiplierBeyond customer-facing features, Gusto uses AI internally for massive productivity gains:Customer Support Superpowers: Their support team uses internal AI tools to access information more quickly, reduce copy-pasting between documents, and pre-populate responses.24/7 Self-Service: AI has made their help center “dramatically more efficient, more effective, better at getting you the information you need.”Document Processing Automation: For tax filings, insurance enrollment, and other complex processes involving third-party documents, AI handles categorization and resolution. “Today, over half the tax notices that come in to the team get automatically flagged, organized, categorized, and then resolved without a human being involved whatsoever.”The Business Impact: More Than Just Cool TechnologyReeves tied everything back to business fundamentals:Increased Operating Margin: Better customer experience and internal productivity improvements create more operating margin to reinvest in new products and improvements.Sustainable Growth: “Gusto’s been free cash flow positive for several years. We think if you’re at the scale we’re at, running a good business means you generate revenue, you generate free cash flow, you reinvest that, you build better product, you serve the customer better, and this technology basically adds to that flywheel.”The Technology Tailwind PhilosophyReeves closed with a broader perspective that every SaaS founder should internalize: “The building blocks constantly are changing.” When Gusto started 14 years ago, their technology tailwinds were cloud, paperless, and mobile. Their go-to-market tailwinds were social and search.“AI is now a part of that and there’s going to be more in the future. And so the mindset here as a company builder is how do you not predict the future because no one has a crystal ball. But how do you stay on top of obsessively what the trends are, what the different technologies are.”The key is matching technology capabilities with real customer problems: “It’s not using technology for the sake of it. It’s you’re looking at a customer problem, a customer pain point… And then if you think that technology can be a way for you to go make that pain go away to improve the quality of your experience, that’s a fantastic reason to dive in.”Top 3 Mistakes to Avoid (Based on Gusto’s Experience):* Don’t Try to Plan Too Far Ahead – Gusto learned that AI roadmaps can’t extend beyond 90 days because the technology and learnings evolve so rapidly. Traditional annual planning will leave you building the wrong things.* Don’t Assume Conversational is Always Better – While conversational interfaces are powerful, not every task is better handled conversationally. Many things are still more effective with traditional graphical interfaces.* Don’t Apply Traditional SaaS Development Processes to AI Projects – Heavy project tracking, rigid engineering processes, and traditional development rituals can kill the rapid experimentation needed for successful AI product development.Thanks for reading The Secrets To Scaling in The Age of AI! Subscribe for free to receive new posts and support my work. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

The latest 20VC x SaaStr episode with Harry Stebbings, Jason Lemkin, and Rory O’Driscoll is here! The team is discussing Databricks’ $100B valuation, the coming IPO tsunami, CoreWeave’s massive debt raise, AI infrastructure spending that could hit trillions, the return of SPACs as a bubble signal, and why AI tool consolidation will happen faster than anyone expectsBottom Line Up FrontHarry Stebbings: “We’re seeing the biggest wealth transfer in tech history unfolding before our eyes. When Databricks hits $100B and Andreessen could make $30B on a single deal, we’re not just in a bubble—we’re in a generational moment where the next wave of IPOs could make 2021 look like the appetizer.”Jason Lemkin: “I think we’re all going to live in AI 24/7 and use 10 times the tokens and 10 times the compute in 24 months. The math is staggering—if we need 200x more infrastructure, how do we even finance that? But here’s the thing: we’ve already automated 5 humans with 10 AI agents at our small team. This isn’t theoretical anymore.”Rory O’Driscoll: “All this depends on 3-5 more years of continued AI capex expansion. If that’s the case, everything works. If not, all bets are off. We’re way out on the risk curve, and the only thing between us and Armageddon is AI adoption continuing at this pace.”Databricks Hits $100B: The New Normal for Private ValuationsWhen Databricks crossed the $100 billion valuation threshold this week, the most telling reaction wasn’t celebration—it was a collective shrug. As Jason noted, “If you’d said 5 years ago there’s going to be a hundred billion dollar market cap private company, you’d be like no way. Now you’ve got Anthropic at $170B, SpaceX at $360B, OpenAI at $500B. The correct response is yeah, whatever.”But beneath this seemingly blasé attitude lies a fundamental shift in how we value AI infrastructure companies. Databricks is now worth ~50% more than Snowflake while growing 2x as fast—crossing $4 billion ARR at 50% growth versus Snowflake’s $4 billion at 26% growth. At 25x revenue, it actually feels undervalued given the growth trajectory and AI positioning.The real story here isn’t just another unicorn—it’s about generational wealth creation concentrated in private markets. If Andreessen Horowitz led Databricks’ seed round in 2013 with follow-on investments across multiple funds, they could own 15% of a company heading toward a $200B IPO. That’s potentially $30 billion in returns from a single deal.“If A16Z invested out of a $650 million fund at the time and they’ve turned a billion and a half into $30 billion, you’re going to feel good in the morning,” Rory observed. “The one thing we don’t take enough into account is just how many people make money when these go out. There are so many LPs in SPVs, SPVs on SPVs. There are dentists who are going to make 5x.”The Coming IPO Tsunami: Why This Is Just the BeginningWhat makes this moment particularly significant is that we haven’t seen the epic IPOs yet. CoreWeave, Circle, Hinge Health, even Figma—these are still “niche plays” compared to what’s coming. As Harry pointed out, “The mega ones are still to come. We might get the next wave of IPOs that are even better very quickly, and the amount of froth that could create in the ecosystem could create a bubble on top of a bubble.”Consider the pipeline: If Canva IPOs next year at $4 billion ARR growing 40% (versus Figma at $1 billion growing 48%), and if Databricks goes public at an implied $200B valuation, we’re looking at a wealth creation event that dwarfs anything we’ve seen before.“There’s a significant absolute sum locked up in private markets that at some point is going to seek a public market where a number of them will be worth $50 billion, which is unprecedented. A few will be worth $100 billion and maybe one will be worth a trillion dollars in the private markets,” Rory noted. That trillion-dollar private company prediction isn’t hyperbole—it’s math.CoreWeave’s $11B Debt Bet: Canary in the Coal Mine or Smart Capital Allocation?While everyone celebrates Databricks’ equity valuation, CoreWeave’s $11.2 billion debt raise tells a different story about AI infrastructure financing. The company spent $22 billion in capex this year while generating $1 billion in quarterly revenue—a capital intensity that would make even utilities blush.“Once you decide to spend $22 billion in capex, you’re going to have $11 billion in debt,” Rory explained. “The way you think about CoreWeave is they’re the guys doing the pointy end of the bet that Meta, OpenAI, Anthropic are talking about—we want to deploy a ton of GPUs and data centers. CoreWeave is stepping up as the financing vehicle.”But this creates a fascinating risk dynamic. CoreWeave becomes the canary in the coal mine for AI demand. “You’re not going to see it in Microsoft. OpenAI isn’t public. You’re not going to see it in Google—they’re too big. But you could see that canary in the coal mine in this public one due to exposure,” Jason observed.The company’s debt structure depends entirely on matching long-term customer commitments with long-term financing. If Microsoft and OpenAI have seven-year take-or-pay contracts and CoreWeave has seven-year debt, the math works. But if customers start paying rather than taking, or if new data center construction slows, CoreWeave stock could signal broader AI infrastructure headwinds before they show up in hyperscaler earnings.The Return of SPACs: Peak Bubble Signal?Nothing says “bubble” quite like Chamath Palihapitiya announcing a new SPAC with the disclaimer that if investors lose everything, they should “embody the adage from President Trump that there can be no crying in the casino.”“It does seem to remarkably correlate—the return of Chamath to the SPAC market appears to correlate with the most recent bubble in 2021 where it worked in the bubble and didn’t work after,” Rory noted. But the bigger issue isn’t timing—it’s structure. SPACs create adverse selection because promoters get paid when deals get done, regardless of quality.“If you pay people to do deals, deals get done regardless of quality,” Rory explained. “I just think it’s an imperfect vehicle for fundraising, and the fact that it’s back is indicative of bubbly times.”The casino metaphor is particularly galling because capital markets aren’t supposed to be zero-sum games. “The point of Wall Street is to funnel money from savers into profitable long-term investments. It’s not a freaking casino, which is a zero-sum game where only the house wins.”AI Labor Replacement: The Trillion-Dollar QuestionSam Altman’s prediction that AI infrastructure will require “trillions” in spending sounds hyperbolic until you consider the scale of potential labor replacement. Jason’s team has already replaced 5 humans with 10 AI agents, and their AI tool costs are approaching $500,000 annually—heading toward $1 million.“I think we’re all going to live in AI 24/7 and use 10 times the tokens and 10 times the compute in 24 months,” Jason predicted. “If we’re using AI 20 times more a day and 10 times more tokens, how much more do we need to spend on infrastructure? What’s 200 times today?”But Rory pushes back on the trillion-dollar timeline: “The $365 billion being invested by the four most profitable companies on the planet is already straining their balance sheets. We’re reaching the limits of what people want to finance.” Even if the direction is right, the pace matters enormously when you’re talking about 4-5% cost of capital on hundreds of billions in investment.The key insight is that AI adoption needs to transition from technology budgets to human labor budgets to justify trillion-dollar infrastructure spending. “If you see that transition, you’ve got $10 trillion of value unlocked, and then spending trillions becomes a lot more feasible,” Harry noted.The Consolidation Wave Is Coming Faster Than ExpectedOne of the most practical insights from the discussion is that AI tool consolidation will happen much faster than the SaaS consolidation cycle. When Jason’s small team is already spending $500,000 annually on AI agents, with every B2B AI startup trying to charge $60-100K per year, the math becomes unsustainable quickly.“We may see a wave of consolidation in AI even next year because it’s just too many AIs that are six figures and up,” Jason predicted. “Everyone thinks their thing is going to save labor, but you can’t all get credit for the same labor.”This creates a different dynamic than traditional SaaS, where companies could afford to have dozens of specialized tools. In AI, you might want everything from one vendor in 12 months, which favors companies like Rippling that can bundle multiple AI agents with an orchestration layer.The implication for venture investing is profound: “We’re ignoring so many risks in AI B2B. Platform risk is everywhere, and we’ve given up worrying about it because the growth is so attractive.”Fintech’s Global Winners: The Nubank StoryWhile everyone focuses on AI, Nubank’s $2.5 billion quarterly profit (up 42% year-over-year) with 123 million customers shows what winning looks like in fintech. The Brazilian neobank is now worth $60 billion and has essentially become the primary bank for an entire generation.“New Bank has 35% market share of Gen Z in Brazil, and Revolut has almost 20% of everyone in the UK with a Revolut account,” Jason noted. This isn’t just fintech disruption—it’s complete market capture.The success pattern is consistent across geographies: find the weakest, most inefficient part of incumbent banking infrastructure and attack it relentlessly. Nubank went full-stack banking in Latin America, Revolut started with FX in Europe, and Chime focused on deposit arbitrage in the US. The bigger the opportunity (weaker the incumbents), the bigger the outcome.Revolut’s product expansion strategy is particularly instructive: 26 new products in development, run like internal venture teams with separate funding goals and weekly testing metrics. As Harry observed, “Out of all the founders I’ve interviewed—hundreds and thousands—Nick [Storonsky] is the best I’ve ever met at this playbook.”Key Takeaways* Private market valuations aren’t crazy when you adjust for growth rates and AI positioning—Databricks at 25x revenue growing 50% actually looks reasonable compared to public comps* The next 12-18 months could see the biggest IPO wave in history, with companies like Canva, Databricks, and Stripe potentially going public at unprecedented valuations* CoreWeave’s debt levels make it the leading indicator for AI infrastructure demand—watch their quarterly results for early signals of market shifts* AI tool consolidation will happen much faster than SaaS consolidation due to budget constraints and the overlap in labor-saving claims* Trillion-dollar AI infrastructure spending is directionally correct but timeline matters—the pace of adoption versus capital deployment costs will determine winners and losers* Successful fintech companies capture entire demographic cohorts, not just market share—35% of Gen Z isn’t just usage, it’s generational platform lock-in* We’re all living in an AI bubble whether we realize it or not—QQQ and VTI exposure means everyone is betting on continued AI infrastructure expansionQuotable MomentsOn the IPO pipeline: “The mega ones are still to come. We might get the next wave of IPOs that are even better very quickly, and the amount of froth that could create in the ecosystem could create a bubble on top of a bubble.”—Harry StebbingsOn bubble dynamics: “If you’d seen a $100 billion private company valuation a year ago, I would have fallen out of my chair. Now it’s funny how I don’t blink an eye in the age of AI.” —Jason LemkinOn wealth creation: “When Andreessen could make $30 billion on the Databricks IPO, how does that further reinforce the cycle we’re in? 30 billion is a lot.” —Harry StebbingsOn AI infrastructure: “CoreWeave are the guys doing the pointy end of the bet that Meta, OpenAI, Anthropic are talking about. They’re stepping up as the financing vehicle to make that happen.” —Rory O’DriscollOn consolidation timing: “We may see a lot of these AI folks get consolidated out sooner than happened in the SaaS wave. Everyone wants an agent for their department. They want five or six agents.” —Jason LemkinOn market risk: “We’re way out on the risk curve, every one of us, right now. The only thing between us and Armageddon is AI adoption. If AI adoption keeps happening, this wagon train keeps rolling. If it slows down, it’s going to get ugly real fast.” —Rory O’DriscollOn fintech disruption: “New Bank is old bank. If the CEO of Bank of America went to Brazil today, he’d be like, ‘I know exactly what these guys are doing’—banking the middle class, top to bottom.” —Jason LemkinOn the trillion-dollar question: “Sam Altman said they’re going to spend trillions on infrastructure. Do you think that’s a real number or a metaphor? No company on the planet has spent a trillion dollars in capex ever.” —Rory O’DriscollThanks for reading The Secrets To Scaling in The Age of AI! Subscribe for free to receive new posts and support my work. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

At SaaStr's packed AI Summit 2025, Box CEO Aaron Levie and IBM VP of AI Raj Datta did a deep dive together with SaaStr's Jason Lemkin on how B2B companies should think about AI agents. With 10,000 attendees—a massive jump from last year—the energy around AI agents was electric.Here are the top 10 learnings from their convo:1. AI Agents Represent a Fundamental Shift from Software to Digital LaborThe Key Insight: We've moved beyond chat interfaces to AI that actually performs work autonomously.Levie explained the evolution: "We've had AI models for five-plus years. Then we had assistants like ChatGPT. But now we're seeing agents that fundamentally go do work for you—and that work could take a minute, an hour, or 100 hours for the agent."For B2B companies, this changes everything. Instead of selling software to 10 lawyers in a company, you're now selling "infinite legal capacity." IBM proved this works at scale, saving $3.5 billion internally through AI agents handling HR and procurement functions.Takeaway: Start thinking about your software as digital labor, not just tools. What work can your AI agents do autonomously for customers?2. Your Customer Data Is Your Biggest Competitive AdvantageThe Key Insight: The companies with the richest, most proprietary datasets will win in the AI agent era."What data are you sitting on that is proprietary to you?" Levie asked. "Very quickly you realize more companies are actually in the data business than they initially thought."Box exemplifies this perfectly. Customers who've stored documents for years can now ask complex queries like "Tell me everything where I have the wrong indemnity provision" or "What contracts I shouldn't have signed." This transforms static data into dynamic business intelligence.Takeaway: Audit your proprietary data assets. What unique insights could AI agents extract that your competitors can't replicate?3. Enterprise AI Adoption Is Happening 1000x Faster Than CloudThe Key Insight: Unlike cloud adoption, which took over a decade, AI is being embraced immediately by enterprises."It's going 1000x faster than cloud adoption because everyone's using it," Datta noted. While pitching cloud to banks in 2008-2009 was a "non-starter," today there isn't an enterprise that doesn't already have an AI strategy in development.The speed difference is staggering: ChatGPT reached 500 million users in roughly two years—faster than any technology in history.Takeaway: Your enterprise customers are ready to buy AI solutions now. The question isn't if they'll adopt AI, but which vendor they'll choose.4. IBM's Agent Catalog Could Revolutionize B2B DistributionThe Key Insight: Even small B2B startups can now access enterprise sales channels through IBM's new agent marketplace."Even if you're a five-person shop, you submit your AI agent to our catalog, and IBM sellers are able to sell it for you," Datta explained. This democratizes enterprise sales in unprecedented ways.IBM's massive sales force essentially becomes the distribution channel for thousands of specialized AI agents, potentially creating the "App Store moment" for enterprise software.Takeaway: Explore partnerships with larger platforms that could distribute your AI agents. Distribution partnerships may be more valuable than ever.5. The Next Generation Will Completely Reshape B2B WorkThe Key Insight: AI-native workers entering the workforce will challenge every existing business process.A Stanford student approached the moderator saying "20% of my class has already dropped out to do AI" after building an AI sales agent that reached $2 million ARR in months.Levie predicted: "They're going to come into our enterprises and say, 'You take 2 weeks to come up with a marketing plan? That was just generated by Claude in 5 seconds. Why would you meet about that?'"Takeaway: Prepare for a workforce that expects AI-first processes. Your internal tools and customer solutions need to match this expectation.6. Data Assets May Soon Appear on Corporate Balance SheetsThe Key Insight: We're heading toward a fundamental revaluation of what makes B2B companies valuable."Right now there's nothing on a company's balance sheet that approximates the value of their data," Levie observed. "In 10 years, will we see 'how good is your data?' emerge as a measurable business asset?"This suggests enterprise valuation models will need to account for data quality, uniqueness, and AI-readiness.Takeaway: Start thinking about your data as a balance sheet asset. Clean, organized, proprietary data will become increasingly valuable.7. You Have a 2-Year Window to Win or Lose EverythingThe Key Insight: The AI transition window is much shorter than cloud, creating extreme risk and opportunity."You could lose your entire position probably in a 2-year period," Levie warned, "but it's also where you could cement your leadership position in a very unique way."Unlike cloud, where companies had years to decide, AI advantages will compound quickly. The winners will get locked in fast.Takeaway: Move aggressively now. Waiting 6-12 months could mean losing your market position permanently.8. Model Context Protocol (MCP) Strengthens Rather Than Threatens MoatsThe Key Insight: Making your data more accessible actually increases customer stickiness.When questioned about whether MCP would weaken competitive advantages, Levie argued the opposite: "The more places where customers can use their data is only a net positive. We want to connect to every MCP environment possible."This mirrors how APIs strengthened platforms by creating more integration points, not weakening them.Takeaway: Embrace interoperability standards. They'll make your platform more valuable, not less defensible.9. AI Agents Work Best When Customers Don't Know They ExistThe Key Insight: The best B2B AI implementations are invisible to end users.Datta emphasized: "There'll be thousands of agents working in the background, but typically one interface for the client. You don't know if you're talking to SuccessFactors or other products—you're just talking to the Watsonx bot."Customers want outcomes, not technology. They don't care about having "a thousand agents"—they want their problems solved seamlessly.Takeaway: Hide the complexity. Focus on delivering better outcomes, not showcasing your AI architecture.10. Traditional Business Fundamentals Now Matter More Than EverThe Key Insight: As AI capabilities become commoditized, execution and customer focus become the primary differentiators."The things that transcend AI will become even more important," Levie concluded. "Building a company, hiring a sales team, iterating, building better products, serving customers—the lowering of barriers to build doesn't make it easier to build sustainable businesses."While AI democratizes software creation, it doesn't make building great companies any easier. Long-term thinking, customer obsession, and execution excellence remain paramount.Takeaway: Don't let AI distract from business fundamentals. The companies that win will combine great AI with great execution.The Bottom Line for B2B Leaders: AI agents aren't just the next feature to add—they represent a fundamental shift to selling digital labor instead of software tools. The companies that recognize this shift and act on it in the next 24 months will cement dominant positions in their markets. Those that don't risk becoming irrelevant.The race is on. Your customers are ready to buy AI solutions now. The question is whether you'll be the vendor they choose, or whether a competitor—potentially one that didn't exist six months ago—will capture that opportunity instead.Thanks for reading The Secrets To Scaling in The Age of AI! Subscribe for free to receive new posts and support my work. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

How We Built ChatGPT Enterprise's Sales Team from Absolute Zero: The Complete Playbook with Maggie Hott, GTM Leadership at OpenAIAbout Maggie: Maggie Hott has spent 15 years building go-to-market teams at four unicorns that collectively represent over $50B in market value. She started as the 2nd SDR at Eventbrite, became the first sales hire at Slack (helping scale from $50M to $1B ARR and a $27B Salesforce acquisition), served as Director of Sales at Webflow (scaling from $40M to $140M ARR), and now leads go-to-market at OpenAI where she built ChatGPT Enterprise from scratch. She also runs a venture fund with seven other women investors, backing 30+ founders. These are her personal views, not those of OpenAI.It was early 2023. OpenAI had just launched ChatGPT, the fastest-growing consumer app the world had ever seen. We were riding an incredible wave, but we had a critical hypothesis: ChatGPT Enterprise would require a fundamentally different go-to-market motion than our API business.When OpenAI hired me to build ChatGPT Enterprise from scratch, I walked into what can only be described as a beautiful blank slate—and a terrifying challenge.Our entire sales and go-to-market organization was less than 10 people. No SDRs. No solution consultants. No customer success managers. No sales operations. No RevOps. No marketing enablement. We didn't even have a working Salesforce instance.What we did have: six incredibly talented account directors and one technical success partner, all laser-focused on selling our API to developers and technical teams.Here's the complete playbook for how we built what we believe became the fastest-growing enterprise application in history.The Build vs. Adapt Strategic DecisionMost companies in our position would have taken the "efficient" approach: enable the existing team to sell both products, maybe hire a few specialists, and gradually expand capabilities.We chose the opposite path: build a dedicated ChatGPT Enterprise team from absolute zero.This wasn't just about headcount. It was about creating an entirely separate organizational DNA optimized for enterprise selling.Why We Built Separate vs. Adapted ExistingProduct Complexity Was Fundamentally Different* API sales required deep technical conversations about integrations, rate limits, and model parameters* ChatGPT Enterprise needed business impact discussions about productivity, compliance, and organizational change management* The buyer personas didn't overlap—CTOs vs. CHROs, CFOs, and business unit leadersSales Cycles Had Different Rhythms* API deals often moved quickly with technical evaluation periods* Enterprise required lengthy security reviews, compliance discussions, and change management planning* Different stakeholders, different timelines, different objection patternsGo-to-Market Motions Required Different Muscles* API was largely product-led with sales-assist for larger accounts* Enterprise needed traditional enterprise selling: demos, pilots, RFP responses, and executive alignmentSpeed Trumped Efficiency in This Moment The market window was massive but narrow. Every week mattered. Having a dedicated team meant:* No competing priorities or split focus* Ability to move at startup speed even within a scaling company* Clear ownership and accountability for outcomesPhase 1: Foundation Building (Months 1-3)The First Three Critical HiresEnterprise Account Executive #1: Financial Services Specialist* 8+ years selling enterprise software to banks and insurance companies* Deep understanding of compliance requirements (SOX, PCI, etc.)* Existing relationships with CISOs and risk management teams* Experience with 12+ month sales cycles and complex procurement processesEnterprise Account Executive #2: Technology Sector Specialist* Background selling to high-growth tech companies* Understanding of developer tools and technical infrastructure* Experience with both startup buyers and enterprise technology teams* Ability to bridge technical and business conversationsEnterprise Account Executive #3: Healthcare/Life Sciences Specialist* Healthcare technology sales background* HIPAA compliance expertise* Relationships with healthcare CIOs and innovation teams* Understanding of clinical workflow integration challengesWhy Vertical Specialists First? Enterprise buyers expect deep industry knowledge. They want to know you understand their specific compliance requirements, regulatory challenges, and business context. Hiring generalists would have slowed our credibility-building process by months.Building the Foundational SystemsCustomer Qualification Framework We couldn't use our API qualification criteria. Enterprise buyers had different needs:* Company Size: 1,000+ employees (later expanded down to 500+)* Budget Authority: Direct access to decision-makers with budget* Use Case Clarity: Specific productivity or efficiency goals* Security Readiness: Existing enterprise software deployment experience* Timeline: Willingness to run pilots and structured evaluation processesInitial Pricing and Packaging Structure Started with three tiers based on our early customer research:* Starter: Small teams, basic enterprise features* Business: Department-wide deployment, advanced security* Enterprise: Organization-wide, full compliance and customizationBasic Sales Process Framework* Discovery Call: Understand use case, stakeholders, and decision process* Technical Demo: Customized demo showing specific business scenarios* Security Review: Deep dive on compliance, data handling, and enterprise requirements* Pilot Program: Structured 30-60 day pilot with clear success metrics* Business Case Development: ROI modeling and executive presentation* Contract Negotiation: Enterprise terms, security requirements, implementation planningThe Customer Success FoundationWhy Customer Success from Day One Enterprise deals aren't won when the contract is signed—they're won when the customer is successfully deployed and seeing value. We hired our first Customer Success Manager in month two.First CSM Profile* Enterprise software implementation experience* Change management background* Technical enough to understand AI/ML concepts* Business-focused enough to measure productivity impactPhase 2: Systems and Process Scaling (Months 4-9)Adding the SDR EngineWhy SDRs for Enterprise? Enterprise deals require extensive relationship building and multi-threading. Our AEs needed to focus on advancing qualified opportunities, not prospecting.First SDR Hires (3 people)* Industry Focus: Each SDR aligned with our AE verticals (Financial Services, Technology, Healthcare)* Enterprise Experience: All had experience prospecting into large organizations* Multi-Threading Skills: Ability to identify and engage multiple stakeholdersSDR Success Metrics* Qualified meetings booked (not just meetings)* Multi-stakeholder engagement (average 2.3 contacts per account)* Account penetration depth (director level and above)Implementing Proper Sales OperationsThe Salesforce Build-Out This was bigger than just CRM implementation. We needed:* Custom Objects: Enterprise-specific fields for compliance requirements, use cases, stakeholder mapping* Workflow Automation: Lead routing, opportunity progression, approval processes* Integration Stack: Marketing automation, customer success platforms, billing systems* Reporting Infrastructure: Pipeline forecasting, activity tracking, conversion metricsSales Operations Hire* Background: Enterprise SaaS sales operations experience* Technical Skills: Salesforce administration, data analysis, process automation* Strategic Thinking: Ability to design scalable processes for rapid growthBuilding Enablement and Competitive IntelligenceSales Enablement Program* Industry Training: Deep dives on financial services, healthcare, and technology sector challenges* Competitive Intelligence: Detailed battlecards for Microsoft, Google, and emerging AI competitors* Demo Certification: Standardized demo flows for different use cases and industries* Objection Handling: Frameworks for common enterprise concerns (security, compliance, change management)Content Creation* Industry-Specific Case Studies: Early customer success stories by vertical* ROI Calculators: Productivity impact modeling tools* Security and Compliance Documentation: Detailed technical specifications for enterprise buyers* Executive Briefing Materials: Board-ready presentations on AI strategy and implementationPhase 3: Full Enterprise Motion (Months 10-18)Scaling the Core TeamAdditional AE Hires* Expanded to 12 enterprise AEs across verticals* Added government/public sector specialist* Hired retail/e-commerce focused rep* Added manufacturing/industrial specialistSolution Engineering Team* Technical Specialists: Deep AI/ML expertise for technical evaluations* Industry Solutions Engineers: Vertical-specific implementation expertise* Demo Engineers: Specialized in creating compelling enterprise demonstrationsExpanded Customer Success* Implementation Specialists: Focused on enterprise deployment and change management* Strategic CSMs: Relationship management for largest accounts* Technical Success Engineers: Post-deployment optimization and advanced use case developmentChannel Partnerships and EcosystemSystem Integrator Partnerships* Deloitte: Enterprise AI strategy and implementation services* Accenture: Large-scale transformation projects* IBM: Hybrid cloud and enterprise integration* PwC: Risk management and compliance implementationTechnology Partnerships* Microsoft: Azure integration and enterprise infrastructure* Salesforce: CRM integration and workflow automation* ServiceNow: IT service management and workflow optimization* Slack/Teams: Productivity platform integrationsMarketing Engine DevelopmentAccount-Based Marketing Program* Target Account Lists: 500 highest-value prospects per vertical* Personalized Campaigns: Industry-specific content and outreach* Executive Events: CIO roundtables and AI strategy workshops* Webinar Series: Industry-specific use case demonstrationsContent Marketing Strategy* Industry Reports: AI adoption trends by vertical* Executive Whitepapers: Strategic guides for AI implementation* Customer Success Stories: Detailed case studies with ROI metrics* Thought Leadership: Speaking engagements at industry conferencesPhase 4: The Great Integration (Months 19-24)The Bold Organizational DecisionBy early 2024, we faced a new challenge: many customers wanted both API and ChatGPT Enterprise capabilities. Having separate teams was creating customer confusion and internal inefficiencies.We made a dramatic decision: unify both organizations into a single go-to-market team.The Integration ProcessThe Scope: 500 people across both teams The Challenge: Most people had been at OpenAI less than 6 months The Change: Everyone got new roles, new managers, new workflows, and had to learn the opposite productIntegration Timeline* Week 1: Announced the change and new organizational structure* Week 2-4: Individual role assignments and team restructuring* Month 2: Cross-training program launch (API sellers learning ChatGPT Enterprise, ChatGPT Enterprise sellers learning API)* Month 3: New process implementation and system integration* Month 4-6: Performance optimization and culture integrationThe Results* Faster Execution: Eliminated duplicate processes and conflicting priorities* Better Customer Experience: One team, one relationship, unified solution selling* Reduced Costs: Eliminated redundant systems and overlapping functions* Enhanced Capabilities: Every seller could now handle both simple API needs and complex enterprise requirementsIntegration Lessons LearnedWhat Worked* Clear Communication: Transparent about the why behind the change* Fast Timeline: Ripped the band-aid off quickly rather than prolonged transition* Investment in Training: Intensive cross-product education program* Leadership Alignment: United leadership team modeling the integrationWhat Was Challenging* Learning Curve: Everyone had to master new products and processes simultaneously* Cultural Integration: Merging two different team cultures and ways of working* System Complexity: Integrating different CRM instances and operational tools* Customer Communication: Managing customer relationships during the transitionThe Key Success FactorsHire for Chaos, Not ComfortEvery early hire was a "chaos translator"—someone who thrived in ambiguous, rapidly changing environments. We specifically avoided people who needed clear processes and defined roles. Instead, we found builders who could create structure from nothing.Maintain Extremely High Hiring StandardsDespite the pressure to scale quickly, we never compromised on quality. One exceptional enterprise seller was worth three average ones, especially in the early days when they were helping define our entire go-to-market approach.Invest in Industry Expertise EarlyGeneric enterprise selling skills weren't enough. The investment in vertical specialists from day one paid massive dividends in credibility, deal velocity, and win rates.Build for Scale from the BeginningEven with a small team, we designed processes and systems that could handle 10x growth. This meant more upfront investment but prevented costly rebuilds later.Embrace Bold Organizational ChangesThe integration decision was risky and painful, but it was the right move for customers and the business. In AI, the pace of change requires constant organizational evolution.The Numbers Behind the SuccessWhile I can't share specific revenue figures, here are some metrics that demonstrate the impact:* Team Growth: 10 to 500 people in 24 months* Win Rate: Maintained >60% win rate even during rapid scaling* Sales Cycle: Average enterprise deal closed in 4-6 months (fast for enterprise AI)* Deal Size: Average contract value increased 5x through pilot program optimization* Customer Success: >95% of enterprise customers expanded usage within first year* Integration Success: Post-integration, unified win rates increased 15% over separate teamsWhat I'd Do DifferentlyStart with Sales Operations EarlierWe should have hired sales operations in month one, not month four. The operational debt we accumulated in early months took significant effort to clean up.Invest More in Change ManagementEnterprise AI adoption requires significant organizational change management. We should have built change management expertise into our customer success function earlier.Build Competitive Intelligence FasterThe enterprise AI landscape evolved incredibly quickly. We should have invested in systematic competitive intelligence gathering and analysis from day one.Create More Structured OnboardingWith rapid hiring, our onboarding process became inconsistent. A more structured program would have reduced time-to-productivity for new hires.The Broader Lessons for Building Enterprise TeamsProduct-Market Fit Looks Different in EnterpriseConsumer product-market fit is about engagement and retention. Enterprise product-market fit is about business impact and organizational change. The metrics and success criteria are fundamentally different.Enterprise Buyers Want to Buy from ExpertsGeneric enterprise selling doesn't work in specialized markets like AI. Industry expertise and technical depth are table stakes, not differentiators.Speed Matters More Than PerfectionIn rapidly evolving markets, the team that moves fastest often wins, even if their processes aren't perfect. Iteration speed trumps initial optimization.Culture Scales Through People, Not ProcessesOur best cultural decisions were hiring decisions. The right people created the right culture, which then influenced all our processes and systems.Customer Success is Revenue, Not CostIn enterprise software, customer success directly drives expansion revenue, renewal rates, and reference-ability. It's not a support function—it's a growth engine.Building ChatGPT Enterprise from zero taught me that enterprise go-to-market isn't just scaled-up SMB selling. It requires different people, different processes, different systems, and different organizational structures.The most important lesson: in fast-moving markets like AI, organizational agility matters more than organizational perfection. The teams that can build, scale, and adapt quickly will win.Thanks for reading The Secrets To Scaling in The Age of AI! Subscribe for free to receive new posts and support my work. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com