It’s on most board agendas now, usually somewhere before lunch. AI update, twenty minutes. And what comes back is a list. A copywriting tool in marketing. A chatbot trial in customer service. A licence count, a pilot “showing promise”, a slide with logos on it. Everyone nods, the item is minuted, and the business is officially doing something about AI. What nobody in the room can say is what any of it is for.
The request itself is entirely rational. In one widely cited survey, 79 per cent of CEOs said they feared losing their jobs for moving too slowly on AI. Non-executives read the same headlines everyone else does, and investors now ask about AI in the same breath as cash. The board isn’t being lazy when it asks for an update. It is managing anxiety with the only instrument the agenda offers. The trouble is that an update measures activity, and activity is the wrong unit.
The forecast, not the dataset
Years ago I stood in front of senior rooms presenting consent and compliance data, and learned a lesson that has outlasted the subject matter. Boards do not want the meteorology. They want the weather forecast. Not the pressure systems and the isobars, but whether it is safe to sail on Thursday. The detail exists to be interrogated when the forecast is doubted, not to be presented first.
AI updates, as currently delivered, are meteorology. Model names, tool inventories, pilot counts, adoption percentages. What the board actually wants to know is the forecast: are we going to be faster and more decisive than the brands we compete with, and what would have to be true for the answer to be yes? I haven’t yet seen an AI update that answers that question, or one that knows it has been asked.
The gap between the two is not hypothetical. When leading commerce technology vendors were asked to rate their own retail customers’ AI readiness, the average score was 4.4 out of 10, and in a separate study of brand manufacturers, not a single one expressed full confidence in their data foundations. An activity list hides this beautifully. You can run nine pilots on foundations that couldn’t support one production system, and the update will look terrific right up until the spend doesn’t.
What to ask instead
Three questions do more work than “what are we doing with AI”. What do we know: which numbers would this board act on without a caveat slide? Where’s the pattern: the one place where intelligence would compound an advantage we already hold, rather than decorating a weakness? And what’s the play: the single deployment we would fund properly, measure honestly, and stop cleanly if it fails? A board that asks those three is having a strategy conversation. A board that asks for an update is collecting a list.
None of this is an argument for slowing down. Clarity is not the homework you do instead of moving fast. It is the reason some brands are moving faster than you.
The next time the agenda says AI update, try striking the item and replacing it with a question: what would we point this at, and would we trust what it reads? Twenty minutes on that will move more than a quarter’s worth of pilot inventories. The board doesn’t need to understand the technology, and it never did. It needs to recognise its own business in the plan. That isn’t an update. That’s the job.