Sparks & Signals: An AI-Powered Audio TLDR

Most frontierAI research isn't built with retail in mind, but a handful of models in Microsoft Foundry Labs map onto real retail problems remarkably well. I went through the catalog and picked out the ones with a genuine use case, from product imagery and conversational commerce to search, optimization, and 3D shopping. I also dig into why in-house retail teams can treat these as building blocks rather than off-the-shelf products, and how pairing them with startups in the Microsoft for Startups portfolio can accelerate the work. If you're a retailer or brand thinking about where to build, I'd love to hear from you.

I spent the last month building autonomous projects: a hashtag#retail operation that designs and sells t-shirts off trending phrases, a pipeline that turns photos into 3D-printed figurines, and a newsletter that translates retail news into insights and actions rather than just headlines.I learned a lot along the way. That you no longer need coding skills to build real things, and that you can build them in a fraction of the time. That debugging can be handed to hashtag#AI, and watching it explain its reasoning is genuinely educational. That the hard part has quietly moved from writing code to saying clearly what you actually want.But the biggest realization was this. All of this technology is democratizing the ability to create. The challenge no longer lives in the tools or the skill to code something. It lives in creativity, in coming up with ideas worth building that no one else has thought of. That is the new skill all of us need to develop, and it is what becomes our unique strength.

What if your news feed answered "𝒘𝒉𝒂𝒕 𝒅𝒐𝒆𝒔 𝒕𝒉𝒊𝒔 𝒎𝒆𝒂𝒏 𝒇𝒐𝒓 𝒎𝒆" instead of "what happened"? I wrote a job description for an #AI agent using #MicrosoftScout, that reads 30+ #retail news sources every week and delivers trends and takeaways rather than headlines. New post on how I built it, the exact instructions that power it, and a link to the first live issue.

Earlier today I had a great conversation with Georges F. Mirza, creator of the ARS² Framework and founder of ComTask, and it prompted me to write down something I keep seeing with the #retailers and #brands I work with.Around 70,000 B2B Tech #startups launch every year and most fail within two years. Yet the capabilities retailers need most right now, especially in agentic AI, live inside those young companies. The result is a familiar cycle: a promising demo, a quick assessment, a pilot that stalls, and the search starts over.In this article I lay out two parallel paths out of that cycle. The first is the curated route through Microsoft for Startups and the Pegasus program, where solution categories are defined through direct customer engagement, sourced from tier 1 #VC portfolios, and vetted for enterprise readiness. The second is for the startups you find yourself, at trade shows or through your own scouting, where two frameworks let you do the vetting independently: ARS² to confirm the solution delivers accurately, repeatedly, at scale, and with speed, and the Azure Well-Architected Framework to confirm the engineering underneath can carry it.Thanks to Georges for the conversation that sparked this. The full article is linked below, and I'd welcome perspectives from others who evaluate startups for enterprise deployment. What does your process look like?

In this issue I dig into the trend I find most significant right now, an agentic layer that overlays existing mature systems instead of replacing them, with great examples from Auger, Intelo.ai, Buynomics, and SimpliContract. I also explain MCP servers in plain terms (think USB ports for AI) and highlight Microsoft for Startups Pegasus portfolio #startups like Nimble, Omnistream, YDISTRI, Tembi - Market Intelligence, and Toolio that are shipping them today. Plus a look at my own experiments running autonomous retail stores with #AI agents.

Merchandise planning is one of those domains where deep vertical focus wins. Toolio built their platform around the actual vocabulary of #retail planners, things like open-to-buy, size curves, and weeks of supply, with apparel brands like Bombas, Knix, and AKA Brands on the customer roster. The results speak for themselves: AKA Brands cut SKU count by 50 to 75 percent while holding sales steady.What I find most interesting is their MCP Server. Across the hashtag#startups I work with, I keep seeing the same pattern: products are becoming capabilities that a customer's own AI agents can call on. Toolio exposes its entire planning brain through #MCP, so a #retailer's agents in Copilot Studio, Claude, or ChatGPT can query plans, run exception searches, and act on inventory data with the same permissions and governance the planning team already has.

What happens when you let AI agents run an entire retail store?I just published my second experiment in autonomous retail. The first was a store that scanned trending news and auto-designed viral t-shirts. This one is Mini Me: upload a photo, and a pipeline of agents turns it into a 3D-printed figurine of you, a loved one, or your pet.Both were really experiments to teach myself how to build and coordinate agents by running something real instead of a toy demo. The biggest lesson? The breakthrough came when I stopped making my agents rebuild everything from scratch and started treating them as operators that drive existing tools, like having an agent run Meshy AI for the 3D models and Printfield for printing and shipping.

If you're building a B2B tech startup in retail or CPG, getting in front of the right enterprise teams is often harder than building the product itself. This piece looks at the corporate innovation programs and CVCs that are genuinely set up to work with external tech providers: from PepsiCo Labs and LVMH to General Mills 301 INC and Tesco Labs, and what it actually takes to convert a pilot into a lasting partnership.

Most supplychain compliance failures are not caused by careless companies. They are caused by a visibility problem that most brands do not realize they have until a shipment gets detained at the border. This article is about how that plays out in practice and what a new generation of trade intelligence tools is doing about it. Trademo is one of the startups we are working with through Microsoft for Startups Pegasus portfolio that is building genuinely interesting solutions for Retail and CPG companies in this space.

Excited to welcome Buynomics to the Microsoft for Startups Pegasus portfolio. After a fantastic conversation with Ingo Reinhardt and Tim Schneider, it's clear they've built something genuinely differentiated in the 𝑹𝒆𝒗𝒆𝒏𝒖𝒆 𝑮𝒓𝒐𝒘𝒕𝒉 𝑴𝒂𝒏𝒂𝒈𝒆𝒎𝒆𝒏𝒕 space. Their agent-based simulation approach gives commercial teams a digital twin of their customer ecosystem to stress-test pricing, promotions, and portfolio decisions before they reach the market.

A recent conversation with my old friend Audrey Piquemal got me thinking about how enterprises are approaching workforce transformation in the #AI era. Merci, Audrey! Too often, companies respond to change by replacing experienced employees with newer and cheaper talent. It may improve short-term financials, but it also strips away institutional knowledge, customer context, and operational judgment that take years to build.As AI takes over more routine execution, those human capabilities may become even more valuable. I wrote about why the future may belong to organizations that focus less on replacement and more on reinventing the talent they already have, along with the emerging platforms trying to make that possible.

Every July, Microsoft's fiscal year turns over and I sit down to think seriously about where to focus my energy for the year ahead. For FY27, one of my five priority areas is 𝑨𝒈𝒆𝒏𝒕𝒊𝒄 𝑹𝒆𝒊𝒏𝒗𝒆𝒏𝒕𝒊𝒐𝒏, and I have written up my thinking on what it means specifically for Retail and CPG enterprise software.𝐓𝐡𝐞 𝐬𝐡𝐨𝐫𝐭 𝐯𝐞𝐫𝐬𝐢𝐨𝐧: the first wave of enterprise software remembered. The second wave explained. The third wave acts. Agentic AI is not arriving as a set of incremental improvements to the platforms that have run retail and CPG operations for the past two decades. It is arriving as an autonomous orchestration layer that runs the workflows those platforms currently support, while leaving the systems themselves intact.That last point is the one I spend the most time on with CIOs. This is not about ripping out your ERP or your WMS. It is about giving them an autonomous brain. The years of custom configuration, the edge-case workflow rules, the compliance logic built up through live operational experience, all of that becomes the skill library that agents invoke. The institutional knowledge stays. The human bottleneck in the middle of the workflow goes.

For years, my passion has been in the world of IoT and building connected systems that can sense, interpret, and respond to the world around them.As I’ve spent more time building AI agents, I kept noticing something familiar: The architectural DNA behind agentic AI looks remarkably similar to the systems many of us built in IoT.

Most retail analytics platforms are built around queries. You form a hypothesis, build a report, get an answer. The limitation is that the most valuable insights are often hiding in questions nobody thought to ask. This is a spotlight on Microsoft partner DataGenie, and how their approach to proactive, continuous analytics changes that equation for retail and CPG teams, and why the difference between finding an insight on Monday versus three weeks later matters more than most organizations realize.

After publishing my piece on letting AI agents run a retail store, the most common question I received: how does it actually work? This post answers that in full technical detail.Three agents, three different foundation models selected for task fit rather than convenience, one sequential pipeline orchestrated through 𝑮𝒊𝒕𝑯𝒖𝒃 𝑪𝒐𝒑𝒊𝒍𝒐𝒕 𝑪𝑳𝑰, and a live storefront on RedBubble that sources, designs, and publishes its own products without manual intervention. I cover the agent configurations, the prompt engineering, the Playwright automation, the pipeline wiring, and the honest account of what needed iteration before the output was any good.I also get into two agents I am planning to add next: one for intellectual property and cultural responsibility screening, and one for automated multi-channel marketing across Instagram, Facebook, and beyond.

I spent some time recently building something I have been curious about for a while: a retail store run almost entirely by AI agents. No manual design work, no product research, no uploading listings by hand. Just a three-agent pipeline that scouts trending content, generates original artwork, and publishes finished products to a live storefront.The store is real & running. Read about how it works and what it revealed about where agentic AI is heading for retail. I Let Agents Run a Retail Store.

Supplychain software has a workflow problem, not a capability problem. Planners spend their days navigating dashboards, reconciling signals across systems, and manually triaging exceptions. The tools work. The work just never stops feeling like tool management.Enmovil, the newest member of the Microsoft for Startups Pegasus portfolio, is taking a different approach. Their CADDIE platform sits as an orchestration layer across existing ERP, TMS, and WMS infrastructure and brings autonomous decisioning directly into Microsoft Teams. The goal, as founder Ravi Bulusu put it in our conversation this morning, is not to replace what enterprises have built. It is to make the intelligence invisible.Steve Jobs said the best technology disappears. Enmovil is building toward that for supplychain logistics.

Enterprise AI is moving beyond copilots and dashboards toward systems that participate directly in how work gets done. In operationally intensive industries such as distribution, manufacturing, and field services, this move is particularly visible. These environments rely on complex workflows, fragmented systems, and deep domain expertise. Traditional automation has struggled to keep up with that complexity.InstaLILY AI (a Microsoft for Startups Pegasus partner) represents a new approach to this problem. Rather than layering intelligence on top of existing tools, the platform introduces AI teammates that operate within enterprise systems and execute work end to end. These AI teammates are designed to function alongside human teams, extending their capacity while working within the constraints of real-world operational systems.

Retailmerchandising is one of the most mature areas in enterprise software, yet much of the work still relies on disconnected systems and manual coordination.I explore how Intelo.ai approaches merchandising as a continuous, agent-driven system that connects planning, allocation, and in-season optimization.If you are exploring how AI can reshape retail planning and execution, this is a space worth paying attention to.

Agentic AI is starting to show up in places we often overlook. Contract management is one of them.In this piece, I explore how SimpliContract is turning contracts into a continuous layer of intelligence across the enterprise.This came out of a great conversation with Guru Venkatesan, CEO and Founder of SimpliContract, on financial efficiency, execution, and the untapped value sitting inside agreements.

I have been spending time looking at startups that rethink how work gets done using AI agents. Auger clearly stands out.Instead of adding another layer of dashboards, it connects data directly to action. The result is a supplychain that can sense changes, evaluate tradeoffs, and execute in real time.Here is my perspective on what this means and why it matters.

My experiment setting up Openclaw on a Raspberry Pi

A recent conversation with Paula Macaggi | OFFBounds stayed with me. She pointed me to an article from Block on the idea of moving from hierarchy to intelligence. We had been discussing Openclaw and what persistent, always-on systems might mean for how organizations operate. The article introduced a set of concepts that added depth to something I had been thinking about for a while.

Retail Stores are moving toward real-time awareness.A look at how movement, shelf interaction, and operations come together to drive action, featuring Ariadne, ShopperAI, and Worlds

Over the past year, I have spent a great deal of time with retailers and consumer brands across different markets. There is real progress being made with AI, data, and new ways of operating, but there is also a sense that these capabilities are still developing in separate layers.What many teams are looking for is a way to connect how the business understands the market, how it responds to customers, and how it executes across stores and channels.That is the context behind an invite-only Retail and CPG Startup Showcase and Networking Event we are hosting in Paris on April 23.

Au cours de l’année écoulée, j’ai passé beaucoup de temps avec des enseignes de retail et des marques de grande consommation sur différents marchés. Des avancées concrètes sont réalisées autour de l’IA, de la donnée et de nouvelles façons d’opérer, mais ces capacités continuent souvent d’évoluer de manière cloisonnée.De nombreuses équipes cherchent aujourd’hui à mieux relier la compréhension du marché, la manière dont elles répondent aux attentes des clients, et l’exécution en magasin comme sur l’ensemble des canaux. C’est dans ce contexte que nous organisons, le 23 avril à Paris, un événement sur invitation dédié au Retail et au CPG, réunissant un Startup Showcase et une session de networking.

Retail complexity is often described in terms of scale. In practice, it is about keeping pricing, inventory, and product data aligned across systems. After a conversation with Maria J Marti (CEO & Founder of ZeroError.ai), I spent some time reflecting on how data consistency is becoming foundational, especially as we move toward agentic workflows.

A few weeks ago, I was in a conversation with a brand that has been investing steadily in digital transformation. Their question was simple and familiar: Which startups would you recommend for a company that still carries legacy systems but wants to move into Agentic AI?It is a question that often gets framed as a tooling decision, though it tends to unfold differently in practice. The answer sits across three dimensions that need to move together. Architectural guidance reshapes the foundation, mature partner solutions provide stability and integration, and curated startups introduce new capabilities through innovative technologies while accelerating specific parts of the journey.

Retail operations break down at the moment you don’t know what’s happening at the store. Power status is one of the most fundamental signals, yet it has been one of the hardest to access in real time.I’ve been looking at how Gisual is changing that.From instant power visibility to triggering the right workflows and even accelerating insurance recovery, this is about turning a simple signal into operational action. Sharing a few thoughts on what this means for Retail Store IQ.

Marketing teams are not short on data. They are short on systems that act on it. I have been spending time thinking about how we move from insight to execution, and what it takes to compress that cycle from days to hours.Sharing a perspective on how agentic ecosystems, enabled through Microsoft for Startups and Pegasus partners like A.Team, are starting to reshape how marketing actually operates.

Spent some time thinking about where agentic systems actually slow down in retail. It is not intelligence. It is trust.Most teams are still in human-in-the-loop mode. That works until scale forces a shift to policy-driven execution.This is where things get interesting. Sharing some thoughts on how blockchain, smartcontracts, and identity could play a role as a trust layer for autonomous systems.

NVIDIA GTC was a different kind of experience for me. Coming from Retail and CPG events, this was a chance to step into AI-first conversations. What stood out was not just what AI can do, but what it takes to make it work in real environments. A few thoughts from the week, including conversations with startups and enterprise teams.

I kept hearing the same terms used interchangeably. Agent. Orchestrator. Claw. So, I spent some time digging in to better articulate what each one actually means and how they fit together in real systems (& in the retail world). A big takeaway for me is that it is a shift from coordination to continuous operation.Agents execute. Orchestrators coordinate. Claws run the system.Sharing my take. Would value your perspective.

Annual reports contain far more than financial results. They reveal the strategic priorities and operational challenges enterprises are working to solve.In this article, I share the approach I use to analyze annual reports, break down those priorities into specific capabilities, and map them to innovative startups using the Business Model Canvas.It is a simple framework that helps connect enterprise strategy with startup innovation in a practical way.

After a few days at MWC 2026 in Barcelona, one observation stood out.𝑹𝒐𝒃𝒐𝒕𝒊𝒄𝒔 𝒊𝒔 𝒃𝒆𝒄𝒐𝒎𝒊𝒏𝒈 𝒂𝒏 𝒆𝒏𝒅𝒑𝒐𝒊𝒏𝒕 𝒇𝒐𝒓 𝑨𝑰.For years AI lived behind screens. Chat windows and dashboards. Now sensors, edge compute, and connectivity allow AI to interact directly with the physical world.Autonomous inspection. Robotic logistics. Infrastructure maintenance. Intelligent machines operating across cities and networks. 𝐓𝐞𝐥𝐞𝐜𝐨𝐦 𝐢𝐧𝐟𝐫𝐚𝐬𝐭𝐫𝐮𝐜𝐭𝐮𝐫𝐞 𝐰𝐢𝐥𝐥 𝐩𝐥𝐚𝐲 𝐚 𝐜𝐞𝐧𝐭𝐫𝐚𝐥 𝐫𝐨𝐥𝐞 𝐚𝐬 𝐟𝐥𝐞𝐞𝐭𝐬 𝐨𝐟 𝐫𝐨𝐛𝐨𝐭𝐬 𝐫𝐞𝐥𝐲 𝐨𝐧 𝐜𝐨𝐧𝐧𝐞𝐜𝐭𝐢𝐯𝐢𝐭𝐲, 𝐨𝐫𝐜𝐡𝐞𝐬𝐭𝐫𝐚𝐭𝐢𝐨𝐧, 𝐚𝐧𝐝 𝐝𝐢𝐬𝐭𝐫𝐢𝐛𝐮𝐭𝐞𝐝 𝐜𝐨𝐦𝐩𝐮𝐭𝐞.I share a few thoughts from the show floor.

The telecom industry stands at an inflection point. Networks are evolving into intelligent platforms. Edge compute is moving closer to the source of data. AI is becoming embedded into infrastructure.At the same time, a parallel shift is underway in the physical world. Robots are leaving controlled factory floors and entering dynamic environments across logistics hubs, retail spaces, smart cities, and industrial sites.At MWC 2026 in Barcelona, this convergence takes center stage. And one of the most compelling companies operating at this intersection is General Robotics , joining the Microsoft for Startups delegation at the Microsoft booth.

At Mobile World Congress 2026, connectivity will be everywhere. The real question is what rides on top of it.Helfie.AI offers one such opportunity.Helfie transforms a standard smartphone into an AI powered health screening tool. Using video, image, and signal analysis, it enables users to conduct rapid, accessible health assessments from their own devices. The experience is simple. The implications for telecom are profound.For telecom operators, growth from traditional voice and data has plateaued across many markets. The next wave of differentiation will come from services that deepen daily engagement, increase relevance, and elevate the carrier from utility to platform.

Telecom leaders spend billions building networks for a future they must predict before it arrives. Fiber rollouts. 5G expansion. Edge infrastructure. Device subsidies. Content partnerships. Every major investment depends on one question:Where is digital demand moving next?At Mobile World Congress 2026, Microsoft for Startups is bringing a company that approaches this question from an unexpected angle: That company is Tembi - Market Intelligence.

Modern telecommunications operators face constant operational and customer experience challenges. Growing data volumes create opportunities, but many teams struggle to turn insights into action. Networks are resilient, but customer experience often falters when real-world disruptions occur, especially power outages. A single outage can drive tens of thousands of support inquiries and distress calls, and without fast, accurate context, support teams and customers alike are left guessing.Today’s telcos need a way to diagnose the right root cause rapidly and act immediately to resolve customer issues across channels. Gisual and Notch together provide that capability: real-time, accurate outage intelligence combined with autonomous customer engagement and support automation.This is a new category of experience orchestration, one that moves telcos from reactive support to proactive resolution.

Microsoft for Startups comes to Mobile World Commerce with a clear purpose: to help enterprises turn innovation into execution.Every organization is under pressure to move faster and deliver more with less. Yet many ideas stall between proof of concept and real deployment. Our role is to close that gap.Startups bring speed and bold thinking. Microsoft brings a trusted, global cloud and enterprise-grade security. Together, this creates a co-innovation model that is practical and measurable.Our mission in Barcelona is simple: spark conversations that become partnerships. Not demos for show. Not pilots that stall. But collaborations that move from idea to impact.At MWC, we connect enterprise leaders with startups ready to scale on Microsoft’s platform. These are production-ready solutions built for real systems, real data, and real business constraints.

Specialty retail was built on expertise. But what happens when the expert is not on shift? From plumbing and flooring to wine and beauty, in-store specialized knowledge is becoming one of retail’s scarcest resources. High turnover, complex assortments, and long onboarding cycles make it difficult to scale true category depth across every location. In this blog post, I explore how domain-trained Personal AI models deployed at the edge can help preserve institutional knowledge, accelerate onboarding, and turn every associate into a confident category advisor. If you are thinking about how to protect margin, increase basket size, and elevate in-store experience, this is a conversation worth having.

We often ask customers what they want.But as the classic milkshake story taught us, customers do not buy products. They hire solutions to make progress in their lives. I first encountered the 𝑱𝒐𝒃𝒔 𝒕𝒐 𝑩𝒆 𝑫𝒐𝒏𝒆 𝑭𝒓𝒂𝒎𝒆𝒘𝒐𝒓𝒌 when we were shaping Microsoft for Startups. It changed how I think about founders, enterprises, and innovation. The real question is not what features people prefer. It is what job they are trying to accomplish.In this article, I explore how that thinking evolves in the age of AI, from behavioral consumer intelligence platforms like i-Genie.ai to real-time market visibility from Nimble and Tembi - Market Intelligence, and how Microsoft for Startups helps enterprises find and validate the right startups to accelerate in-house innovation.

I have been tracking a few key technology themes across #CES, #NRF, and #WEF this year. One topic stood out clearly: #AgenticCommerce. This post reflects on the signals from all three events and what they reveal about how autonomous systems are beginning to reshape digital markets.

A few weeks ago, I shared a blog post (https://lnkd.in/gwmg3PbH) exploring what #retail could look like when multiple #AI agents work together rather than in silos. The response was overwhelming and quickly turned into action. The founders of the #startups Uriel Knorovich(Nimble), Wendy Chen(Omnistream), and Mario Megela(YDISTRI) decided to collaborate, alongside Krutika Kamdar from Microsoft, to make the concept real and bring it to life on stage at #NRF2026. In just two weeks, they turned a thought experiment into a working multi-agentic #retail scenario. My sincere appreciation to Krutika for the leadership and energy she brought to driving this effort.

As AI agents take on more execution, people move up the stack, from tasks to outcomes. I wrote this to explore what that shift really means in practice. Also hear about how Microsoft for Startups is using AI Agents.

Retail Store IQ is not about more dashboards. It’s about turning signals into action at the edge of the store. In this piece, I build on my earlier Retail Store IQ article and explore how sentiment, behavior, context, competition, assortment, inventory, and execution come together to shape better in-store experiences.

Over the past year I have been looking for a startup that can help brands make sense of the fast-moving signals that shape consumer interest. Not by looking at last year’s data, but by reading what customers are paying attention to right now across search, social, creators and other public sources. i-Genie.ai has built one of the most compelling approaches I have seen, turning these external signals into clear insight that helps brands stay ahead of emerging trends.

My 2026 outlook on the technologies I will be watching closely in retail. From Agentic Commerce and Spatial Models to RetailStoreIQ and intention understanding, these are the parts of the intelligence stack that I believe will have a meaningful impact on how the industry operates.

There’s a quiet shift happening in grocery. Protein aisles are expanding, meals are shrinking, and shopper intent is changing faster than most realize.In my latest post, I explore how CueZen, a Microsoft for Startups Pegasus partner, is helping retailers understand this new shopper and deliver healthier, more personalized journeys.

Reading a Docker article (https://lnkd.in/g_h8BgiP) about microservices sent me down an unexpected rabbit hole. It made me realize there’s a striking parallel between the hidden complexity of microservices and the emerging world of Agentic AI.