The 90-day pattern
We see the same arc in almost every Copilot rollout we walk into. Month one is full of energy — leadership announcements, a kickoff demo, a flurry of curious prompts. By month two, daily usage has quietly halved. By month three, the licences are still being paid for, but most of them aren't being touched.
The interesting bit is why. It's almost never that Copilot doesn't work. It's that the rollout was treated as a software deployment when it needed to be treated as a behaviour change.
Three things the successful rollouts do differently
1. They start with a workflow, not a feature tour.
The failed rollouts open with a session called "Introduction to Copilot." The successful ones open with a session called "Use Copilot to prepare for your Monday status meeting." Same product, completely different framing — and a 4–6× difference in week-four active use.
2. They make one person on each team the local champion.
Not a power user. Not a tech evangelist. Just one normal person who's been given an hour a week to be slightly ahead of their colleagues and answer questions. Distributed peer support beats a central helpdesk every time.
3. They measure adoption, not licences assigned.
Licences assigned is a vanity metric. Weekly active users by team, prompts per user per week, and self-reported time saved are the metrics that actually predict which teams stick with it.
What this means for your rollout
If you're inside the first 90 days, the most useful thing you can do this week is pick a single high-frequency workflow that hurts — meeting prep, email triage, summarising long documents — and design one focused training session around it. Not Copilot in the abstract. That specific job, made faster.
That's the unlock.