Over 9 in 10 have tried it, fewer than 1 in 4 have made it a habit: what your Copilot report doesn't tell you

In early November, I opened the Copilot usage report of an industrial group I work with. The figure at the top of the page was excellent: more than 9 in 10 people had used Copilot during the month.

It is the kind of figure you present to the executive committee. Everyone congratulates themselves, and the question "does it work?" seems settled.

Then I looked at the detail, person by person, day by day. Fewer than 1 person in 4 used it almost every day.

Both figures are true. They do not answer the same question.

What the report calls an active user

In the Microsoft 365 admin center, the Copilot usage report counts as active any user who deliberately triggered a Copilot feature over the chosen period: 7, 28, 90 or 180 days. Opening the pane does not count, sending a prompt does. The active user rate divides that number by the number of enabled licenses (source: Microsoft Learn, Microsoft 365 Copilot usage report, page updated on 08/18/2026).

A single prompt in four weeks is therefore enough to be active. This rate, often presented as an adoption rate, actually measures trial. It is very useful for spotting those who never started. It does not tell you who has changed the way they work.

The grid I use: four bands over 28 days

To know whether Copilot has become a habit, I count the number of days each person used it over 28 days, which is about 20 working days. Then I sort everyone into four bands.

• Almost daily: 15 days or more. Copilot is part of the working day.
• Weekly: 5 to 14 days. The person uses it, but not yet by reflex.
• Occasional: 1 to 4 days. One try, sometimes two.
• No use: 0 days.

In this group, the breakdown looked like this:

• nearly 1 person in 4 almost daily
• more than 4 in 10 weekly
• about 1 in 4 occasional
• 1 in 10 with no use at all

Read this way, the same report no longer tells the same story. You stop asking whether people use Copilot. You see who has built it into their work, who is still hesitating, and where the lever is.

Two precautions make the grid reliable. Always compare 28-day windows without holidays, otherwise December will make your numbers drop. And keep the same reports from one month to the next: changing the source is like changing the thermometer.

The potential is not where you look for it

Faced with these figures, the first reflex is to deal with those who do not use it. It is rarely the best investment: they are few, and often for good reasons (a role that is barely affected, a job without a screen, resistance that does not disappear after one training session).

The real opportunity is the block of weekly users. They have already taken the step. They know how to open Copilot and get something out of it. What they lack is two or three uses that come back every day in their job: planning their week from their emails, summarizing a missed meeting, drafting the first version of a reply.

Moving part of this group to daily use changes the company's productivity more than convincing the last holdouts.

What we do in practice, month after month

On this program, steering comes down to four actions.

1. Measure every month, on the same window. One usage snapshot before each training theme, and one at the review, with the same grid. The dashboard is not there to reassure the committee: it is there to decide on the next session.
2. Train in waves, with practice. Short sessions by theme (Outlook and Teams, then Excel, then Word and PowerPoint), where everyone redoes the steps on their own workstation. Satisfaction is not enough: we also measure who was able to practice.
3. Identify champions. When employees were asked who wanted to become a Copilot champion, nearly half answered yes or maybe. These people exist in every team. You just need to ask them, then give them a role.
4. Track regularity rather than volume. A user who sends three prompts a day is worth more to the company than a user who sends fifty on a single day.

Locating your organization: five levels not to mix up

This grid corresponds to the third and fourth levels of the model I use to steer an AI rollout:

1. Access: who has a license.
2. Training: who knows how to use it in their job.
3. Usage: who has tried it, and how often.
4. Adoption: who uses it regularly, to the point of making it a habit.
5. Impact: what has changed in the work, measured on specific tasks.

Most steering committees stop at the third level and present it as the fourth. That is where false successes are born, along with 2027 budgets that are hard to defend.

For each level, the question to ask, the indicator, the source in Microsoft 365 and the pitfall to avoid are detailed in the guide Measuring the real adoption of Microsoft 365 Copilot.

The right question

The question is not how many people have tried Copilot. It is how many could no longer do without it.

Open your usage report, export the per-user detail, and count the days. In one hour, you will know where you really stand.

If you want to do it with someone who has done it before, or build the plan that will move your weekly users to daily use, let's talk.