What actually blocks a Copilot rollout

Organisations rarely stall on Copilot because the technology does not work. They stall because a decision nobody owns has not been made, and the programme quietly waits for it.

1. No agreed acceptable-use position

Legal, risk and security each hold a partial view of what employees may enter into an AI tool, and none has been asked to write it down. Without a position, every scenario is escalated individually and the programme moves at the speed of the slowest committee. A one-page statement, agreed once, removes weeks of recurring debate.

2. Permission uncertainty nobody wants to own

Copilot surfaces what a user can already open. When nobody is confident what that includes, the safe response is delay. The fix is narrow: review sharing and membership in the specific locations the pilot scenarios touch, and evidence the result. A full-estate review is rarely fundable and rarely necessary to start.

3. Scenarios described as capabilities, not tasks

Summarise documents is not a scenario; it is a feature. Produce the first draft of a monthly service report from the case export and last month report is a scenario, because you can tell whether it worked and how long it took before. Programmes that cannot express scenarios at task level cannot measure them either.

4. A pilot cohort chosen by seniority

Licences allocated to leadership produce enthusiastic anecdotes and no usable evidence. The cohort should be the people whose daily work matches the scenarios being tested, whatever their grade.

5. No owner after launch

Scenario libraries go stale, content decays, and prompts that worked stop working as documents change. Without a named owner and a small standing budget, usage declines from the third month regardless of how well the launch went.

What to do first

1. Write the acceptable-use position and get it agreed.
2. Review permissions in the locations the pilot will touch, and record the result.
3. Rewrite scenarios at task level, with a baseline for effort today.
4. Choose the cohort by task, not by grade.
5. Name the owner for after launch, and fund them.