The question nobody in the room can answer

There is a moment in most board meetings about AI that everyone recognizes and nobody plans for. A director, usually one of the quieter ones, asks why this initiative was funded and that one was not.

It is a fair question. It is also, in most organizations, unanswerable.

Why it is hard

Not because the decision was bad. Usually the decision was fine. The difficulty is that the reasoning behind it never existed in a durable form. It lived in a conversation, a slide, and the judgement of a few people who were in the room and are now somewhere else.

Three things make this worse than it sounds.

The first is that AI initiatives arrive from different parts of the organization, described in different vocabularies, with wildly different levels of rigour behind the numbers. Comparing them is not like comparing two capital projects. It is like comparing two arguments.

The second is that confidence travels better than evidence. An initiative brought by a persuasive senior sponsor gets a different reception than the same initiative brought by a careful junior one. Everyone knows this happens. Almost nobody has a mechanism that prevents it.

The third is that the reasoning decays. Six months later the people have moved, the deck has been superseded, and the spreadsheet behind the business case is on version eleven. What remains is a decision with no visible basis.

What a good answer looks like

A useful answer to the director’s question has three parts, and it is worth being precise about them.

It says what standard both initiatives were held to. Not that they were both reviewed carefully, which is unfalsifiable, but that the same specific things were asked of each.

It says where they differed on that standard. This is the part that actually answers the question, and it is only possible if the first part is true.

It says who decided and what they were accountable for. Not the committee. A person.

Organizations that can produce that answer are not smarter than the ones that cannot. They wrote it down at the time, in a form that outlives the meeting.

The cost of not having it

The immediate cost is a slightly awkward silence, which everyone survives.

The real cost arrives later, in three places. In an audit, where a figure from last year’s pack has to be substantiated and nobody can find its source. In an incident review, where an AI system did something unexpected and there is genuine ambiguity about whether it was ever out of bounds. And in a raise, where an investor asks how AI investment decisions are governed and the honest answer describes a process that lives in people.

None of those are governance failures in the dramatic sense. They are the ordinary consequence of making good decisions without recording their basis.

That is the gap worth closing, and it is closeable well before anything goes wrong.