HappyRobot

Huge thanks to AI Agent Conference for the opportunity to share what we've been up to at HappyRobot! When asked how things have changed since we first went into production two years ago, my answer was simple: the ambition has grown. At the beginning, companies were trying to create point solutions with AI. A one to one translation of an existing workflow. As trust has built in the technology and in the partner, companies have begun to see AI as a transformational tool. Today, companies are redesigning workflows as they automate them. Agents don't exist with the same design constraints. Think of the ticket created for a support conversation that could have been solved live. Instead of replicating that ticket creation process, one of our customers is leveraging 2 agents working in the background to get the issue solved live. One customer call. No ticket needed. Our mission has always been changing how work is done. We're on the next phase of that journey with our customers!

Agent demos are easy. The hard part is what happens after. I moderated a panel today at the AI Agent Conference with two amazing Microsoft for Startups Pegasus founders, Jonathan Corbin of Maven AGI and Javi Palafox of HappyRobot. They delivered this to a packed room, and it's easy to see why. Between the two of them, they work with hundreds of enterprises. Here are my top 5 takeaways from the conversation: 1/ 𝗧𝗵𝗲 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻 𝗵𝗮𝘀 𝗰𝗵𝗮𝗻𝗴𝗲𝗱 Enterprises have moved past "how do we deploy agents" to asking where specifically it can have the largest impact, how they'll measure value, and what timelines look like. The era of deploying agents because a board said so is over. 2/ 𝗧𝗿𝗮𝗻𝘀𝗳𝗼𝗿𝗺 𝘁𝗵𝗲 𝘄𝗼𝗿𝗸, 𝗱𝗼𝗻'𝘁 𝗷𝘂𝘀𝘁 𝗿𝗲𝗽𝗹𝗮𝗰𝗲 𝘁𝗵𝗲 𝘁𝗮𝘀𝗸𝘀 Early customers wanted like-for-like automation, having agents do what a human did. Mature customers are redesigning how work gets done entirely, not just swap who (or what) does it. 3/ 𝗖𝗼𝗻𝘁𝗲𝘅𝘁 𝗶𝘀 𝘁𝗵𝗲 𝗮𝗿𝗰𝗵𝗶𝘁𝗲𝗰𝘁𝘂𝗿𝗲 𝗽𝗿𝗼𝗯𝗹𝗲𝗺 𝗻𝗼 𝗼𝗻𝗲 𝘁𝗮𝗹𝗸𝘀 𝗮𝗯𝗼𝘂𝘁 Agents need the right knowledge, customer context, permissions, and ability to take action systems. Without that, even the best model cannot deliver enterprise-grade outcomes. 4/ 𝗖𝗵𝗮𝗻𝗴𝗲 𝗺𝗮𝗻𝗮𝗴𝗲𝗺𝗲𝗻𝘁 𝗶𝘀 𝗻𝗼𝘁 𝗼𝗽𝘁𝗶𝗼𝗻𝗮𝗹 The best deployments involve aligning teams early, defining ownership, reducing fear, and making employees part of the transformation. Companies that get stuck are often blocked by internal politics and fear. 5/ 𝗔𝗴𝗲𝗻𝘁 𝘀𝗽𝗿𝗮𝘄𝗹 𝗶𝘀 𝗰𝗼𝗺𝗶𝗻𝗴 Many companies are giving teams tools to spin up agents independently. Without shared context, governance, and clear ownership, that can become chaos fast. The winners will not be companies with the most agents. They'll have the clearest operating model. One thread connects all five: The gap between demo and production is not a technology gap. It's an organizational one. 💭 𝗜'𝗺 𝗰𝘂𝗿𝗶𝗼𝘂𝘀: What's the biggest blocker you're seeing in scaling agents to production at your organization or for your customers? PS: Thanks for your valuable insights, Javier and Jonathan! The audience appreciated all the wisdom bombs you kept dropping 👏 Pablo Palafox Erena Estellés Tatay Caitlin Knapp Lillia Lovell Sami Shalabi Bethany Cordes Carolyn Deng AI Agent Conference Simon Chan

Last week I had the privilege of presenting HappyRobot to 200+ leaders at Deutsche Telekom's T3 Summit 2026 in Munich. These are my takeaways 🧠 Deutsche Telekom's T3 leadership gathering in Munich brought together their top executives to confront what AI really means for how they lead, decide, build, and execute. Two things that we talked about on the panel with Nicole Hörl Muñoz and Peter Lorenz that I keep coming back to: 💡 The AI adoption gap in the enterprise is not technology, but change management. Every enterprise leader I talk to has access to the same models, the same compute, the same vendors. What separates the companies pulling ahead is whether they bring their teams along the journey, giving them a real say in how the organization rethinks the way work gets done. AI transformation "done" to a team has a higher chance to fail. But bring the team along, and you can see magic happen. That's a big part of the game, from what I've seen. 💡 Every enterprise is, at its core, solving a coordination problem. That's why our Platform + Deployments model works. We don't show up with a generic tool, or a point solution, and hope it sticks. We go deep with our customers to uncover the most complex, tangled, non-obvious business problems in the business. The ones no other solution dares to enter, and we solve those. Because that's where the real value is unlocked. The obvious automations are commodities. The hard, messy, cross-functional coordination problems are where enterprises win or lose the next decade. -- What I loved most about T3 was the intellectual honesty in the room. Senior leaders asking hard questions, testing assumptions, debating quite openly, and willing to learn in public together. That kind of posture at the top is rare, and it's exactly the cultural foundation that makes change management possible in the first place. Huge thanks to Jonathan Abrahamson for the invite, to Peter Lorenz for daring to run a live HappyRobot demo onstage with me, and to the Nicole Hörl Muñoz for hosting a fantastic fireside chat. And to the entire Deutsche Telekom leadership team, thank you for the openness, the sharp questions, and the energy 🚀

Last Tuesday I was the only outsider in a room full of Naturgy executives at IESE Business School. I was there to share what we've learned working together: what it actually takes to go from a prototype to production, and why that last stretch is harder than it looks in AI. What made it special was the room itself: These weren't people being briefed on AI. They were the ones building - senior executives with real use cases they'd put together themselves, presenting them openly to each other. Hands on keyboard. That's not something you see every day at that level. That's the only way to go now. AI only hits its real potential when leadership understands its capabilities deeply enough to know what's possible - and how to turn that into something real. That's the crucial shift, and the hardest one to make. Seeing it alive at the top of an IBEX 35 🇪🇸🇪🇸 is something else. People say enterprises aren't adopting AI. Some of them are further along than you'd think. Here you have one of them. Grateful to Terrell Sheridan the for the invite; the best is yet to come!! 🚀

A memorable night in Paris 🇨🇵 We knew there’d be brilliant talent We knew there’d be stellar tech But.. what to expect? You never know with humans.. they're anything but deterministic! And lucky we are as a species.  Because Paris’ finest AI builders are creative. And they built some crazy agents yesterday evening. Some demos we had: 1️⃣ You just got locked at your own door. You are in Miami. Your wife is travelling to HappyRobot’s Builder’s night. So what do you do, you find a locksmith ! ! Antoine, Carlo and Yassine Charai (beast) built an agent who calls locksmiths in Miami. The production-ready app and agents were.. sublime. We received 7 quotes in 3 minutes from actual locksmiths in Miami. Problem solved. 2️⃣ You need to run polls at scale. Claude or ChatGPT ? SadRobots or HappyRobot? So Cobalt's Grégory H. and Ismael scraped everyone’s LinkedIn in a killer UI, and randomly called.. Amaury, who answered all ten questions before Constance and Estelle unleashed 60 concurrent calls on the room. Absolute chaos. 3️⃣ Le Baguette Index, a family project my brother and I are building, which we’ll unveil to the general public very soon. A personal one. 4️⃣ Orus's 5 greats automated some of the most painful - and costly workflows of their insurance business, led by Diane. Compelling demo ! Good laughs, great food, crazy building. Commitment to excellence. Everyone won. And a personal note as I relocate, Paris, I will miss you 🩵