How to Measure Microsoft 365 Copilot Adoption (and Prove It Is Worth It)

Every Copilot rollout reaches the same moment. The licenses are bought, the pilot is over, and someone in leadership asks the question every IT leader dreads: are people actually using it. The honest answer, in most organizations, is that nobody knows yet. They know how many licenses they bought, and they know a few enthusiastic users by name, and the gap between those two facts is where the rollout gets judged.

License count is not adoption. A license that nobody opens is a cost, not a capability, and the only way to tell the difference is to measure what people actually do. Microsoft ships the reports to do it, and they are already in the tools most admins use every day. Here is how to run them, and how to turn the numbers into a story leadership believes.

Step 1: Start with the Copilot usage report in the admin center.

Open the Microsoft 365 admin center, go to Reports, then Usage, then Microsoft Copilot, and open the Usage tab. This is the report that answers the basic question: are the people you licensed actually using Copilot, and where are they using it.

Three numbers matter most. Enabled users are the people who hold a Copilot license in the period. Active users are the licensed people who actually used a Copilot feature they started themselves, and the definition is strict on purpose: opening the Copilot pane does not count, but submitting a prompt does. The active users rate is simply active users divided by enabled users, and it is the first number to watch, because it tells you what share of your investment is doing anything at all.

The report also shows prompts submitted and the average prompts per user, plus adoption by app, so you can see whether people are using Copilot in Teams, Outlook, Word, Excel, PowerPoint, OneNote, Loop, or Copilot Chat. That breakdown matters, because adoption is rarely even. In most rollouts, Teams meeting summaries and Outlook drafting take off first, while Word and Excel lag until someone shows people what Copilot can do there.

You can view the last 7, 28, 90, or 180 days, and a given day of data usually appears within 48 hours. The per-user table lists everyone licensed in the past 180 days, including the people who never used Copilot, with their prompt counts, active days, and last activity date. Usernames are hidden by default, which is worth knowing before you go looking for individual names, because the report is built for trends, not for watching one person. Several admin roles can open it, including Global Administrator, Reports Reader, Teams Administrator, and the AI Administrator role.

Step 2: Open the Copilot Dashboard in Viva Insights for the bigger picture.

The admin center report tells you what happened. The Microsoft Copilot Dashboard in Viva Insights, which lives in the Viva Insights web app, tells you what it means. It organizes everything into four pages: readiness, adoption, impact, and sentiment, which is the story executives actually want to hear, in that order.

The dashboard turns on with as little as one Copilot license, and the fuller feature set, including benchmarks, group-level views, and survey-based sentiment, unlocks at 50 or more licenses. Data processing takes up to seven days after the first license is assigned, so do not expect numbers on day one. The dashboard always reflects the previous 28 days, with up to a six-day delay, and it refreshes daily.

The adoption page groups your licensed users by habit, and the definitions are worth learning. Microsoft calls someone a power user when they average 15 or more Copilot actions a week and were active in at least 9 of the past 12 weeks. Habitual users average 1 to 14 actions a week over the same 9-week span, and novices used Copilot in fewer than 9 of the 12 weeks. Those bands are more useful than a single active user count, because they show whether usage is deepening or just flickering. A rollout with a hundred active users and three power users is a rollout that needs training, not more licenses.

The dashboard also tracks Copilot Chat usage by people who do not hold a Copilot license, which is a useful signal on its own. Those are the people already reaching for AI inside your boundary, and they are the most obvious candidates for the next round of licenses.

Step 3: Treat assisted hours as an estimate, not a measurement.

The impact page of the dashboard shows Copilot assisted hours, which is Microsoft's estimate of the time employees were assisted by Copilot over 28 days, and Copilot assisted value, which multiplies those hours by an average hourly rate to produce a dollar figure. Both numbers look precise, and both need a warning label.

Assisted hours are modeled, not measured. Microsoft credits roughly six minutes per search, summary, or creation action, based on its own research with knowledge workers, and meeting summaries count the meeting duration. Nobody timed your employees with a stopwatch, and Microsoft says plainly that the figure is a general estimate. Present it to leadership as a directional number, never as measured time saved, and set the hourly rate to one that matches your own organization instead of accepting the default.

There is one more reason to be careful with the history. Microsoft confirmed that the email draft and email summary metrics, and therefore assisted hours and assisted value, were underestimated from June 1, 2025 through February 18, 2026, and fixed from February 18 onward. If you are comparing this quarter to last year, check whether your data crosses that window before you draw a conclusion from it.

Step 4: Ask people directly, because usage data cannot tell you why.

Numbers show what people did. They never show why they stopped, or what they tried and gave up on. That is why the dashboard includes a sentiment page, and why a short user survey belongs in every adoption review.

Keep it to three or four questions: how much time Copilot saved you this week, whether the output was good enough to use without heavy editing, and what task you most wish it handled better. The answers will tell you more than the usage report, because low usage usually signals a missing scenario or missing training, not a bad tool. If the survey says people do not know what to ask Copilot for, the fix is role-based training with real examples from their own work, not another email announcing the rollout.

Step 5: Pick a few real tasks and measure them.

Usage and sentiment are the first two levels. The third is business outcomes, and it is the one that makes the case for renewal. Pick two or three tasks that matter to your organization, set a baseline for how long they take today, and measure again after people have used Copilot for a few weeks. Meeting follow-up time, proposal turnaround, and time spent drafting routine email are good starting points, because they are common, they are easy to time, and Copilot touches them directly.

The baseline is what makes the later claim credible. If you did not measure before the rollout, start now anyway, and compare the departments that received role-based training with the ones that did not. That comparison is usually the most persuasive number in the whole review, because it shows leadership exactly what training buys.

Step 6: Act on what you find, every month.

A measurement nobody acts on is a report, not a program. Review the numbers monthly, and do three things with them. First, reassign licenses from people who have not used Copilot in 90 days to the Copilot Chat users who are already active, because the readiness report in the admin center flags the heaviest unlicensed users each week, and those are the people most likely to use a license well. Second, aim training at the groups with the lowest active user rate, and use your power users as the trainers, because a colleague showing a real workflow beats a generic session every time. Third, report three numbers to leadership and nothing more: the active users rate, the average prompts per user, and one business outcome with its baseline. Six to eight weeks of a well-supported pilot gives you enough data to decide on expansion, and a monthly review after that keeps the rollout honest.

Copilot will not prove its own value. The licenses arrive, the novelty fades, and the only thing standing between a successful rollout and a quiet renewal question is whether someone is watching the numbers and acting on them. Run the reports, ask the people, measure a real task, and report the three numbers. That is how a rollout stops being a purchase and starts being a program.

Kyle Hollis