Microsoft bills Copilot per seat, every month, whether the person behind it opens it or not. Most companies bought a batch of seats at once, then watched usage spread unevenly. A few people use it daily. Some tried it once. Some never clicked the icon.
The good news: you don’t need a new tool to find out who’s who. The Microsoft 365 admin center already tracks it.
Start with the Copilot usage report
In the Microsoft 365 admin center, go to Reports > Usage and open the Microsoft 365 Copilot report. It lists every licensed user and the last date they used Copilot in each app: Teams, Outlook, Word, Excel, PowerPoint, OneNote, Loop, and Copilot Chat.
Set the time range to the longest period available, then export the user table to CSV. That export is your working file for everything below.
Turn off concealed names first (if they’re on)
If the report shows strings of random characters instead of names, your tenant is hiding user details in reports. That’s a privacy setting, and it’s on by default in many tenants.
A Global Administrator can change it under Settings > Org settings > Services > Reports. Uncheck the option to display concealed user, group, and site names. This changes every usage report in the tenant, so clear it with whoever owns privacy or HR policy before you flip it.
Sort everyone into three groups
Once you have names and dates, every licensed user lands in one of three buckets:
| Group | What the report shows | What it usually means |
|---|---|---|
| Active | Activity in the last 30 days | Keep the seat |
| Tried and stopped | Activity, but not in the last 30 days | Something got in the way |
| Never used | No activity date at all | The seat was assigned by job title, not by need |
Copilot usage report
Every licensed seat lands in one of three groups
Sorted by each user's last Copilot activity date
Active
Used in the last 30 days
Tried and stopped
Some activity, none in 30 days
Never used
No activity date at all
Illustrative example for a 20-seat organization. Your split comes from your own report.
The “never used” group is the obvious savings. The “tried and stopped” group is the more useful one, because those people wanted it to work.
“Last activity” tells you who, not how much
One prompt in Outlook counts as active. So a user who summarized a single email three weeks ago looks the same as someone who drafts every proposal in Copilot.
If you need depth, the Microsoft Purview audit log records Copilot interactions, and you can count them per user. It takes an audit log reader role and some scripting. Practical 365 published a walkthrough that combines the usage report with audit data to score each license. For most small businesses, the three buckets above are enough to make a decision.
Talk to people before you pull seats
When we look at low Copilot usage, the cause is rarely “this person doesn’t need AI.” It’s usually one of these:
- Nobody showed them what to use it for in their actual job.
- The files they need live somewhere Copilot can’t reach, or in a mess Copilot can’t make sense of.
- They tried it on a real task, got a weak answer, and stopped.
A 10-minute conversation with the “tried and stopped” group tells you whether to retrain, reassign, or cancel.
Check permissions before you reassign
Copilot can surface anything the user already has access to. That’s how it’s designed. If a seat moves from a sales rep to someone in finance, Copilot will happily summarize every overshared SharePoint folder that person can open.
Before you hand a seat to someone new, check what they can reach. Overshared sites and “everyone” links are the usual problem.
Then reassign or reduce
Reassigning is simple: remove the license from one user and assign it to another in Users > Active users. Reducing your seat count depends on your contract. Annual commitments often can’t drop mid-term, so check your terms (or ask whoever sold you the licenses) before you plan around the savings.
Our take: review Copilot seats every quarter, and assign them by use case, not by org chart. A seat should follow a specific job someone wants done faster. If you’d rather have someone pull the report and sort it with you, talk to one of our engineers.