Around mid-year, the same thing happens in a lot of Microsoft 365 environments I work with. A Copilot licence review suddenly shows up on the agenda, usually because someone has noticed that not everyone with a licence actually uses it. Copilot isn't cheap, so there's pressure to find inactive users, take their licences back and give them to teams that get something out of them. In my experience, doing this well has little to do with technical tricks and a lot to do with process. You need usage data from Microsoft 365, an allocation model people can follow, and direct feedback from the departments. Put those three together and the review becomes much easier.

The Copilot usage report is the most useful review tool in the admin centre

I always start a Copilot licence review with the Microsoft 365 Copilot usage report. You'll find it in the Microsoft 365 admin center under Reports → Usage → Microsoft 365 Copilot. It shows who has used Copilot in the last 30, 60 or 90 days, split by app (Word, Teams, Outlook, Excel), with a 'last activity date' for each user. That level of detail helps a lot. Some people use Copilot in Teams all the time, while others tried it once in Word and never came back. With the report you can tell real adoption apart from a bit of curiosity.

There is one practical catch. Depending on your org privacy settings (Settings → Org settings → Reports), user names may be pseudonymised. For an actual review you need real names, so you'll have to switch privacy mode off temporarily. Check that you have the right admin role, and talk to your privacy officer before you change it.

flowchart TD
    Licence["Copilot licence assigned"] --> Usage["Copilot usage"]
    Usage --> Report["Copilot usage report"]
    Report --> Feedback["Business feedback"]
    Report --> Decision["Licence decision"]
    Feedback --> Decision
    Decision --> GroupAssign["Group-based assignment"]
    Decision --> Fallback["Copilot Chat fallback"]

Figure: A Copilot licence review that combines usage data with input from the business.

Group-based licensing in Entra makes reallocation much easier

Assigning Copilot licences by hand ends in a mess sooner or later, and you notice it most when it's time to move licences around. Switching to group-based licensing in Microsoft Entra (Entra admin center → Identity → Groups → Licenses) makes it obvious who has what, and rolling licences out or pulling them back becomes simple. Static groups give you tight control, dynamic groups give you automation. Either way, licence management gets easier and much easier to audit. Microsoft's documentation on assigning licences through groups is a good reference for setup and troubleshooting.

This is the workflow I prefer. I start with the usage report and flag users who have had little or no Copilot activity over, say, the past 60 days. They go into a "Low Usage" group. Their licence can then go to someone on the waiting list, and everyone can see how the decision was made. Some organisations tell users before they remove access. That avoids confusion, and you often get honest answers about why Copilot never stuck for them.

Department feedback puts the usage numbers into context

Usage data only tells part of the story. In many tenants I've seen users who rarely open Copilot, not because it's useless to them, but because their role doesn't call for it every day, or because they are still figuring out how to use it well. For that reason I never do a Copilot licence review without input from the business units. Sometimes the low-usage users are early adopters who are still experimenting. Sometimes a department needs more training or clearer use cases. The technical review and the adoption work belong together. If you want ideas for getting more people to use Copilot in practice, take a look at Copilot adoption: scenarios and champions.

Practical tip: I always ask team leads or Copilot champions for a short note on how Copilot helps (or doesn't) in their daily work. That brings out successes and gaps you would never spot in the numbers.

In the end, the licence decision weighs the usage figures against the business context. Who really needs Copilot from here on, and where would another user get more out of it?

A few practical criteria decide who keeps a licence

These criteria have worked well for me in Copilot licence reviews, both for fairness and for ROI:

  • Users with clear, ongoing Copilot activity (for example, 5+ sessions/month)
  • Key roles or multipliers (leaders, IT, internal champions)
  • Departments with defined Copilot use cases or innovation goals
  • Low-usage or inactive users (after checking with their manager)
  • Occasional users: offer them the "Microsoft 365 Copilot Chat" fallback instead

Since January 2025, Microsoft 365 Copilot Chat has been a handy fallback for occasional users. They get conversational AI, but without the deep integration into the Office apps and without awareness of their documents. For many people that's plenty for ad-hoc questions or light automation.

Licence management works best as a recurring cycle

After a lot of Copilot licence reviews, this is what I'd pass on: treat it as a regular cycle, not a one-time clean-up. The usage reports keep improving. Combine them with structured feedback and training, and you can keep expensive licences with the people who actually use them and see a real productivity gain. For more on reporting and day-to-day licence management, see Copilot usage reports: Tracking adoption and license ROI.