Three things to agree before the pilot starts

Pilots rarely fail on capability. They fail because nobody agreed in advance what the agent was allowed to get wrong, so the first visible mistake turns into a referendum on the whole project.

Three things to settle before you build

First, the confidence floor: the score under 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 should move.

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 it still costs money every month.