Skip to content
Gagan Bhatia

Ai4, day 2: governing AI and paying for it

· First published on LinkedIn

Day 2 at Ai4 opened with history.

Watching Geoffrey Hinton, Fei-Fei Li, and Andrew Ng share a stage felt like watching the architects of modern AI debate its future in real time. They disagreed on almost everything - open models, regulation, job. (There was also a lighthearted debate over whether Hinton once worked for Ng at Google Brain. Even AI legends have their "remember when?" moments.)

But for me the thought provoking sessions weren't the headline keynotes, they were the enterprise architecture conversations.

  • S&P Global's Ashit Talukder, in my opinion, was one of the best sessions today. His frame: the shift from "helpful conversations" to accountable execution. Bolt-on AI makes individual tasks faster while the workflow stays slow. Real value starts when agents complete governed work end-to-end: plan, retrieve, decide, act, measure, with human-in-the-loop, evals, and an audit trail. His warning: Shadow IT has become Shadow AI, and teams routing around controls aren't moving faster - they're expanding the risk surface.

  • Panasonic's Elise Neel on rewiring a global enterprise for the AI economy. Two lines I wrote down: "This is the worst version of AI you will ever experience - design for refactoring." And from research she shared on 200 global C-suite leaders: 86% agree AI will fundamentally reshape work, yet fewer than a third believe their own leadership team is aligned on the future.

  • A FinOps talk with a stat I can't unhear: AI spending projected to hit $2.59 TRILLION in 2026 - while most organizations can't tell you what a single token costs them.

  • Red Hat distilled trustworthy agent platforms into four foundational capabilities: isolation, identity, observability, governance.

Three different stages said the quiet part out loud today: the roadmap for enterprise agents runs through identity, permissions, and audit. One of them literally called it a birth certificate for AI agents.

Day 1 was about governing AI. Day 2 was about paying for it. The winners will do both in the same design.

What was your Day 2 highlight?