Main idea

A few days after writing down the tension between the dock and accounting, the Product Manager suggests to the AI making the discrepancy reason optional. In an isolated conversation, the AI would probably have defended the idea and offered a clean screen. In a workspace that contains the two ideas and the tension that connects them, it answers that the proposal reduces the friction described by the logistics managers but runs against financial control's reconciliation need, and that the sources do not yet make it possible to choose between several options.

The AI does not decide in the Product Manager's place. It prevents a new intuition from erasing a reality already established. Its response does not come from general knowledge of product management: it is situated, because it knows which ideas rest on which sources and which contradiction remains open.

Why it matters

The same AI that smoothed things over in a conversation becomes a guardrail against smoothing. What changes is not the model but the material it reads: intact sources and written tensions give it grounds to raise an objection.

Nuances and limits

The guardrail only works for tensions already written down: a contradiction nobody has put into words will not be raised.

Open questions

• How do you keep a guardrail that recalls established tensions from also making it hard to reopen them when a new fact makes them obsolete?