Scoping a pilot that survives month three

Pilots do not usually fail on capability. They fail because nobody agreed in advance what the agent was allowed to get wrong, so the first visible mistake becomes a referendum on the whole project.

Three things to settle before you build

First, the confidence floor: the score below which the agent does not act. Second, the escalation route: exactly which human queue sub-floor work lands in, and who watches it. Third, the review cadence: when you look at the exception log together and decide whether the floor moves.

None of this is technical. All of it is the difference between a pilot that becomes infrastructure and one that quietly stops being used while still costing money every month.