The hardest sentence to say in a product review is that we do not know yet. It is usually the true one.
The situation is familiar enough. Leadership needs a commitment: a date, a number, a claim about what the AI will do for retention. The team's honest answer is a wide range, or an assumption nobody has tested. A wide range is not a good enough answer, and everyone in the room knows it. So the team gives a confident answer instead, and everyone leaves satisfied and no better informed.
I do not think this is a character problem. It is a sensible response to the incentives in the room. Uncertainty, stated plainly, reads as a lack of grip. Confidence reads as competence. People work out quickly which one gets rewarded, and they are not wrong about it.
The playbook I wrote covers the cognitive side of why capable organisations still get build decisions wrong: sunk cost, the reluctance to stop something with momentum behind it, the pull towards evidence that supports a decision already made. This is a different mechanism. It is social rather than cognitive, and it acts on the person speaking rather than the person deciding.
What it costs you is visibility. The weakest assumption in the project stops being discussable, which does not make it any less weak. It gets found later, usually after the money is spent, and often by a customer.
What fixes it is not encouraging people to be braver. It is giving them something to stand behind. Saying you do not know yet is hard in front of a board. Saying that you tested the assumption, here is where it held, here is where it broke, and this is what remains open, is not hard at all. That is a strong position. The difference is not courage, it is evidence.
It reframes what an Evidence Phase is for. On paper it is a short piece of work that tests the assumptions a build decision rests on. In practice, a good part of its value is that it gives a team permission to say the honest thing, with something concrete underneath it.
It also changes what a good answer looks like. Not a narrower range produced by pressure, but a narrower range produced by knowing more. The first is a promise. The second is a forecast.
If a team tells you what you want to hear, it is worth asking what it would have cost them to tell you anything else.
About the author
Dr Paul Sacher is the founder of Sacher AI, a behavioural AI consultancy and product partner for GLP-1 and digital health. He is co-founder and Research Director of the Behavioral AI Institute and an honorary senior lecturer at Imperial College London, with over 26 years across obesity care, behavioural science, and AI.