AI Credits vs Salaries, Token Maxing Goes Corporate

On Unfiltered with Matt & Nige, Episode 10, the boys start recording whilst still laughing, and eventually get into some good chat about spending the same on AI credits as employee salaries 🎙️TLDR: some companies are pushing AI usage so hard that credits are on track with salaries. ServiceNow is going all in on “one front door” plus control. And governance is no longer the boring bit. It is the only way this scales.This week on Unfiltered the two talk all things:* the Speechify founder story, dyslexia, accessibility, and building a massive business off a real need* AI credits are starting to match engineering salaries, and some founders want the credits to exceed the people cost* token maxing and figuring out what to use AI for, and what not to* ServiceNow Knowledge, 27,000 people, Otto, AI Control Tower, and why governance is suddenly front and centre* Duolingo and the displacement question, can AI replace it, or does the streak keep it sticky* and Matt’s mini existential wrestle, building a business vs building actual changeAI credits vs salariesMatt shares a story from 20VC, Harry Stebbings interviewing the Speechify founder. What sticks is not the product, it is the operating model.His engineering team’s AI credits are close to their salaries.And he is pushing for credits to go beyond.That is a proper marker of where we are heading.If you hire someone for 100k, do you assume they will burn 100k of AI credits too.That question is coming for every team.Token maxing, but with standardsNigel admits he is still in the “figure it out properly” phase.He wants to experiment with agents for his newsletter, but he does not want AI to rewrite his thinking. He writes to think. That is the point.Matt relates to that, but also pushes the other side. Founders driving usage hard, selecting best internal use cases, showing the team how it is done.The meta point is simple.Everyone is experimenting.The winners will standardise what good looks like inside teams.Accessibility and the “Enable” storyWho doesn’t love an old story. .Matt talks about his first agency, built around digital accessibility and helping charities and NGOs. And how his mum, to this day, thinks he helps disabled people fix their computers.It is funny, but it is also real.A lot of the most important mainstream tech started as accessibility tech.Speechify sits in that lane.Solve a real need, then it becomes normal for everyone.ServiceNow Knowledge and the control storyNigel gives a quick view from Vegas.27,000 people, massive announcements, and the big themes are consistent.One front door. Otto.Control. AI Control Tower.The message is clear.Agentic is not just “do stuff for me.”It is “do stuff for me safely,” across systems, data, and workflows.Governance is the unlock, not the brake.Duolingo and displacementThey also hit the consumer angle.Duolingo is loved. It is addictive. The streak is real. But the displacement question is also real. AI can translate in real time. You can learn by doing, not tapping an owl for 1,000 days.Matt’s take is nuanced.Duolingo works, but learning style varies, and competitors are everywhere.This is the repeat pattern.Great UX plus habit loops can hold for a long time.Until the interface changes.The Matt wrestle, build a business vs build changeMatt talks about getting pulled into vibe coding a new business idea, then catching himself.The goal is not just “build a thing.”The goal is to create change.It is a founder problem, and it is a good one.Tools make building easy. Purpose keeps you honest.The big takeawayAI is not just a productivity layer now.It is becoming a budget line.Credits. Controls. Governance. Adoption.This is the grown up phase.And if you want impact, the tool is not the mission.The mission is the mission.Subscribe for more Unfiltered every week 🎙️ This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit unfilteredwithmattnige.substack.com