Enablement · 12
Measure one deliverable before you teach the rest
The practical order, for a company at the start, is dull. First, one bounded Cowork deliverable you can compare with last month's version of the same work. Next, a read-only app with a named owner, if Code or Studio is even available to you. Then a Work IQ suggestion checked against the source records. Autopilot, with a spending limit and a person who can step in, comes last, and only for someone in the private preview. That order is a reading of the September staging. It is not a timetable Microsoft published.
What the measure already was
Pick a measure that existed before the project: cycle time on a pack, errors a reviewer marks, a promise that was missed. Do not invent a Copilot score. Microsoft's reported examples are not your baseline.
Teach only what you measured
The enablement content is the gap you saw. If the model missed a figure the reviewer always catches, that figure becomes the exercise. If the pack was fine and the approval was skipped, the exercise is the approval, not another prompt catalogue.
Stop is a result
FrontierOrg, explicitly fictional, allows its fictional board to stop. A real team can do the same in public, inside the company: we tried one deliverable, the check failed, we are not training anyone else yet. That sentence will save more hours than a tour of previews.
Monday
Name the one deliverable and the measure you already have for it. Put every other September feature on a list called 'not this month'.
What result would make you cancel the next training session, and who needs to hear it?
Sources: Microsoft, 16 June 2026, Copilot Cowork is now generally available (opens in a new tab). Microsoft blog, 25 September 2026, Introducing the new Copilot (opens in a new tab).
Where this note mentions FrontierOrg, that is a fictional method for rehearsing a decision. It is not a company, not a client, and not a case study.