Why the pilot is quiet
On announcing less and finishing more.
Most AI pilots we hear about were announced before they started. An email to the whole organisation, a name, sometimes a logo. Then six weeks of one team trying something while everyone else waits for the results, and the results are never quite big enough for the announcement that preceded them.
We run the pilot the other way round. One team, told in person. The tools they already open, with one thing added. A measure they were already tracking before we arrived, so nobody has to invent a number to prove it worked. Adjustments every week, in a twenty-minute conversation, not a review meeting.
Quiet is not secret. The team can talk about it to anyone who asks, and people do ask, because a colleague who has stopped complaining about the weekly summary is more convincing than any email. The difference is that the practice spreads on its own reputation, at the speed people are ready for it, and it arrives with someone who can show how it works at the next desk over.
By week ten there is a policy in plain language, a monthly review that fits in a calendar, and a team that would notice if the thing stopped working. Then we leave. That is the announcement.