AI councils can create clarity. Or they can create theater.
AI councils only help when they clarify decision rights, intake, ownership, and action instead of becoming another governance performance.
I’ve seen both.
On paper, an AI council sounds like the right move. Bring Legal, Security, Product, Data and Engineering together. Review use cases. Align on risk. Make decisions.
But in practice, it can go in two very different directions.
In the worst cases, the council becomes a stage. Slides get presented. Concerns are raised. Notes are taken. And nothing actually moves. Decisions get deferred, ownership is unclear and teams leave the meeting unsure of what just happened.
That’s what I would call AI theater.
In better setups, the council is not there to discuss everything. It’s there to decide the few things that matter.
What makes the difference is surprisingly simple:
- Clear decision rights: who approves what, at what level of risk.
- A defined intake process: not every use case should reach the council.
- A bias for action: if a decision can be made, it should be made in the room.
- Named owners: every AI use case has someone accountable beyond the meeting.
I’ve also learned that smaller is often better. The more people in the room, the easier it is to drift into discussion mode instead of decision mode.
AI governance doesn’t need more meetings. It needs better ones.
And on a Friday, that’s probably the simplest test:
If your AI council disappeared tomorrow, would anything slow down?
If the answer is no, it might be theater.