The committee that has run out of questions

A board that has run out of questions about its AI sub-committee should not pretend otherwise. The committee was assembled in a hurry, between two regulatory shocks and a small but visible miscalculation by a respected peer. Three years on, it has become the most-attended, least-decisive forum on the calendar.

The pattern is recognisable. A board strikes a sub-committee to demonstrate that it is taking a category of risk seriously. The committee draws a clutch of well-credentialled members, a respected chair, and a brief that is broader than it should be. The first year is productive; the second is descriptive; the third is ceremonial.

The sub-committee is the wrong governance answer. It was always going to be the wrong governance answer, because AI is not a category of risk — it is an operating condition. Treating it as a category creates the polite illusion that there is a part of the business that does not have AI in it, and a part that does. Both halves of that proposition are now false.

What replaces the committee is a doctrine. A doctrine has fewer meetings, more accountability, and an author. The author is the chair. The doctrine is read aloud, in plain English, once a quarter at the main board. The committee — if one survives at all — exists to draft the doctrine, not to discharge the responsibility.