Somewhere between the dashboard and the boardroom, something already narrowed your options.

Why It Matters

If you have ever approved a strategic option in a board meeting without asking who or what narrowed it down to that shortlist, this is worth ten minutes before your next meeting, not after a regulator or a shareholder asks the same question for you.

The Take

Board governance was built to scrutinize decisions: who proposed them, who voted, what the rationale was. It was never built to scrutinize what shaped the options in front of the board before anyone spoke. That gap did not matter much when every input came from a named analyst. It matters enormously now that a meaningful share of those inputs are already filtered, ranked, or shaped by a model nobody in the room can name.

The distinction worth holding onto is between two very different kinds of AI in the room. One is visible, used openly for summaries or scenario modeling, and everyone knows it is there. The other is not: embedded inside a dashboard or a recommendation engine, quietly deciding what counts as the top three risks or which options make the shortlist, with no disclosure attached to any of it. Most governance frameworks are only built to catch the first kind.

The Blind Spot

Ask a board member how a new AI tool gets approved and you will usually get a reasonable sounding process: procurement review, security assessment, a signed vendor contract. That process assumes AI arrives as a new purchase. Increasingly, it does not.

The dashboard your finance team already uses, the CRM your sales team already uses, the reporting tool your operations team already uses: all of these regularly ship AI powered features as routine updates to software approved years ago, by a different committee, for entirely different reasons. Nobody reviews a tool for a feature it did not have when it was first purchased, because nobody thinks of a feature update as a new AI tool needing review. The approval question is not hard to answer because the process is broken. It is hard to answer because the process is never triggered in the first place.

The fix is not a bigger review committee. It is a standing rule that any existing vendor shipping a new AI powered feature into a system that feeds board materials gets flagged for disclosure, the same way a material contract change would be, before that feature quietly becomes part of what your board sees.