https://youtu.be/-sw3HD7bDlA?si=YTbAAnLCzWcgTyNx

I just got back from the Arize Observe Agentic AI conference, and here’s the TL;DR:If you think your AI “copilot” is just a smarter autocomplete jammed into a side panel… you’re doing it wrong. This wasn’t an ordinary AI conference. It was raw, practitioner-driven, and full of experiments in building autonomous agents. Which means one thing: The UX stakes just went up. The future of AI isn’t just about model performance. It’s about whether people can actually trust, collaborate with, and understand these increasingly powerful systems. That means designing: 🧠 Agentic affordances (when to intervene, when to let go) 🪞 Digital Twins (representations of users and models, not just dashboards) ✍️ Narratives over metrics (because bar charts won’t earn trust) 🪦 RIP simple side panels (more on that below) I wrote up my sketch-notes and hot takes here: 👉 Read the full post Bonus: Want a free chapter from my book on Digital Twins for AI Product Design? Just comment #UXforAI on this LinkedIn post and DM me — I’ll send it to you. — Next issue: I’ll break down a few real-world Agentic UX patterns I saw in the wild — and which ones are quietly failing behind the scenes. Till next time, Greg 📕 UXforAI.comJoin the conversation