Governance is usually framed as the price of using AI. We would frame it the other way round: clarity is what makes speed possible, because it removes the need to re-litigate the same question for every use case.
Boards we work with that deploy AI confidently tend to have settled five questions once, at the top, and encoded the answers: which decisions may never be delegated to a model; what evidence is required before a model influences a customer outcome; who signs when a model causes harm; how model risk is reported alongside financial and operational risk; and how the board itself stays literate as the technology changes under it.
“A committee that can only advise is a committee everyone eventually routes around.”
Answering those five questions takes a board perhaps three sessions. The return is that the organisation stops treating every use case as a novel ethical question and starts treating most of them as engineering.
The failure mode is delegation without authority. An ethics committee that can only advise becomes a bottleneck that everyone routes around, and the routing around is invisible until something goes wrong. Give the committee a decision right — the power to halt a deployment — and it becomes both faster and safer, because it no longer needs to be consulted on everything to be effective.
The board's real job is not to understand transformer architectures. It is to know which of its decisions are now partly made by software, and to be comfortable with the answer.
