Ai4, day 1: nobody was talking about models
Day 1 at Ai4 was an incredible experience: 12,000 people, and almost nobody was talking about models.
I spent the day in the rooms where AI gets serious — governance, risk, infrastructure, economics. Three threads kept surfacing no matter where I sat:
- Agents are a workforce now. One keynote argued we should stop "governing" AI and start hiring it — budget lines, performance reviews, terminations. An hour later, I was in a session on risk management for autonomous agents. The tension between those two rooms was the most honest conversation of the day.
- Economics is the new architecture constraint. From a former Intel CEO and OpenAI's compute lead calling chips the strategic bottleneck of the AI era, to a session literally titled "We saved the customer $100M in token costs" — cost per outcome is now a design requirement, not a finance afterthought.
- Regulated industries are done waiting. The sessions on governed systems and AI in regulated environments were packed. Those buyers aren't asking "can AI do this?" anymore. They're asking "can I defend this to my auditor?"
After 20 years architecting regulated platforms, here's my take: all three threads are one problem. You can't hire, budget, or audit what you can't identify. Agent identity and permissions are where this whole conversation is heading.
What was your biggest Day 1 takeaway?
