An AI retrospective shows which workflows save time and which create rework. Teams often adopt AI quickly but rarely examine prompts, costs and failure patterns.
Review examples from spaces, credit use, corrections and good prompts to improve rules and assistants.
Start with a narrow boundary: which website, space, file, recipient or decision is affected? This makes the task reviewable instead of turning it into a broad catch-all request.
A useful work order is: “Create an AI retrospective with successes, failure patterns, cost drivers, reusable prompts and improvements.” For important cases, add that uncertainties must be marked visibly instead of being filled in silently.
Pay special attention to risk, review duty, privacy and later findability. These points decide whether the result is only useful for the moment or can be found, checked and continued by the team later.
Do not judge adoption only by impressive isolated examples.
AI usage improves through evidence, not through more random prompting.