The Board That Doesn't Know If It Uses AI

For two years, boards have been asking management for the same thing: an artificial intelligence policy. Which tools are allowed, what data can be used, who approves projects, how risks are measured. It's the right demand. Overseeing AI is now part of every director's duty of care.

But there is one question almost no board asks about itself.

How does the board use AI?

What happens off the agenda

Directors work with AI too. They summarize the board pack, prepare questions, compare contract versions, draft comments for committees. Often they do it with tools nobody approved, using information that should never leave the room.

This isn't speculation. In Deloitte's July 2026 Board Practices Quarterly, 71% of public companies surveyed said they didn't know whether their board or its committees had used AI in meeting-related work. 51% have no AI policy specific to the board. And only 8% require human review or accuracy checks when the board uses it. The sample is small and U.S.-focused, but the direction is clear.

Those who oversee without rules of their own lose the authority to demand them.

Three risks no company policy covers

The first is confidentiality. A director who pastes a strategic report into a public AI tool creates a risk the corporate policy doesn't reach, because that policy was written for employees, not for the board.

The second is accuracy. An AI-generated summary nobody checked can end up in the minutes. And minutes aren't a draft: they are the record of what the board knew and decided.

The third is competence. According to the same study, only 4% of companies include AI competence in director evaluations. What isn't evaluated isn't expected, and what isn't expected doesn't develop. A board that doesn't know what to ask about AI ends up receiving reports instead of governing.

One page, not a manual

The answer doesn't require a long document or a new committee. It requires consistency. A board can write its own AI rules on a single page, signed by every director, answering four questions:

• Do we use AI in our work as a board, and for what?
• What board information never goes into an AI tool?
• Who verifies what AI produces before it reaches the minutes?
• How do we evaluate each director's AI competence?

It's a small amount of work. And it changes the authority with which the board talks to management about AI. From demanding rules to demonstrating them.

Oversight starts at home. A board that governs its own use of AI has something no management presentation can give it: first-hand judgment.

Could your board answer those four questions today without asking anyone?

Bring the blank page to your next meeting. If it's still empty when the meeting ends, that's the first item on the next agenda.