Weeks 3–4: Run and observe

Use real work. Track the primary measure, adoption behavior, error, and guardrail exceptions. Do not expand the test because early results are exciting. Do not hide a disappointing result until the final meeting.