A passing AI pilot still needs an expansion decision

A pilot passing its test does not automatically justify a wider rollout. It answers a narrower question: did this version meet the checks, on this scope, with these inputs?

Try this hypothetical example: an internal document assistant passes an agreed review using one approved source collection. Before adding another collection, keep two records separate:

• Test evidence: which cases were checked, what failed, and what the results do—and don’t—cover.
• Scope decision: who can approve the expansion, which source owners or users need to be consulted, and who will maintain the added material.

If the new collection changes the audience, source quality, or consequences of an answer, treat that as a new decision—not as an automatic extension of the first pass. The decision owner can explain why to expand, defer, or decline, and what evidence would prompt another review.

Which change in scope would make you reopen the decision first: a new audience, new sources, or higher-consequence use?