Most organizations have already bought Copilot licenses. Far fewer can tell you whether those licenses are doing anything.
That's not a reporting problem. It's usually a sign that nobody agreed on what good looks like before the rollout started. Gartner surveyed Microsoft 365 customers and found that only 5% of Copilot pilots had moved on to a broader rollout. Not 5% failed. 5% made it.
We've been running adoption programs since Copilot launched, and that number matches what we see. What surprises people is the reason. It's almost never the technology, and it's almost never the licenses. It's that nobody connected Copilot to the actual work.
of Microsoft 365 Copilot pilots progress to a broader rollout, according to Gartner. Not 5% failed. 5% made it.
The short version
- Licenses assigned tells you what you bought. It says nothing about adoption.
- Monthly active users flatters the number and hides the problem.
- Track three things instead: daily active license rate, prompts per active user, and whether users can name the workflow they use Copilot for.
- Under 15% daily active use isn't a training problem. It's a workflow problem, and more training won't fix it.
- Programs built around real tasks move the number fast. We've seen utilization double inside a month.
Why licenses deployed isn't adoption
Most steering committees get shown one number: licenses assigned. It tells you what you bought. It tells you nothing about what happened next.
The second number people reach for is monthly active users. It's better, but it's still soft. Monthly active means someone opened Copilot once in thirty days. Somebody who tried it on the 3rd, got a mediocre answer and never came back still counts as active for the rest of the month.
We've walked into organizations reporting 70% monthly active where fewer than one in six people had used it that week. Nobody was being dishonest. They were measuring the wrong thing, and the number looked healthy enough that nobody asked a harder question until renewal.
The three numbers worth tracking
Daily active license rate. Of the licenses you're paying for, what percentage were used yesterday? Not this month. Yesterday. It's the only number that survives contact with reality, and it's the one most organizations avoid looking at.
Prompts per active user per month. Volume tells you whether Copilot became a habit or stayed a novelty:
- Four prompts a month means they're experimenting.
- Fifteen means it's occasionally useful.
- Forty or more means they've changed how they work.
Named workflow. Ask ten users to name the specific task they use Copilot for. Not "summarizing" or "writing." The actual task, with the actual document, in the actual week. If they can't name one, you don't have adoption. You have awareness.
That third measure isn't in any dashboard, and it's the most predictive of the three. You can collect it in an afternoon.
What we see in the field
These are real numbers from programs we've run. We've kept the clients anonymous, but the figures haven't been touched.
License utilization sat at 28%. Most of the base had a license and no reason to open it. We ran a twelve week accelerator built around role-based workflows rather than features. Within a month:
The part that matters is that it held. A spike after training is easy. A level that stays up thirty days later is behavior change.
Roughly a third used Copilot daily. A large middle group dabbled. A meaningful share had never opened it at all. One workshop, built around what their people actually do, moved daily use to 79% and eliminated the non-user group entirely. 89% said they'd champion the tool internally. 93% expected to save time every day.
Everyone had licenses. Everyone was using them for meeting notes and email. Nobody was using them for the work that actually pays, which is diligence and analysis. We rebuilt the training around a full deal lifecycle, from data room review through investment committee prep, using a realistic synthetic deal. People stopped asking what Copilot could do and started asking which step to use it on.
Three different industries. Same pattern every time.
A benchmark you can hold yourself against
Be honest about what this is. It comes from our engagement data across enterprise programs in Canada and the United States, not from a survey of a thousand companies. Treat it as a field guide rather than a census. We publish it because the alternative on offer is vendor marketing, and a rough honest benchmark beats a precise invented one.
| Measure | Struggling | Typical | Good |
|---|---|---|---|
| Daily active license rate | Under 15% | 25% to 35% | 60% and above |
| Weekly active license rate | Under 30% | 40% to 55% | 80% and above |
| Prompts per active user, monthly | Under 10 | 15 to 30 | 60 and above |
| Users who can name a workflow | Under 20% | About half | Over 80% |
| Users building their own agents | None | A handful | A visible community |
If your daily active rate is under 15%, you don't have a training problem.
The real problem: workflow, not features
Here's what goes wrong in almost every stalled rollout we get called into.
The organization ran training. Often good training. Somebody stood up and demonstrated what Copilot can do. Summarize a thread. Draft a document. Catch up on a meeting you missed. People nodded. Some tried it that afternoon. Two weeks later usage was back where it started.
Feature training produces demos. Workflow training produces habits.
Look at the difference:
| Feature training | Workflow training |
|---|---|
| "Copilot can summarize email threads" | "Pull the three open commercial issues out of this vendor thread and draft the response for your director to approve" |
| "Copilot can draft documents" | "Turn Friday's site inspection notes into the Monday report your regional manager expects" |
| "Copilot can catch you up on meetings" | "Build the pre-read for Thursday's steering committee from the last three meeting recordings" |
The left column is a party trick. The right column is a task somebody already has on their list this week, and Copilot either makes it faster or it doesn't.
The distinction sounds small. It's the entire difference between 28% and 62%.
When we design a program now, we start from the job rather than the product. What does a credit analyst do on a Tuesday? What does a plant supervisor do at shift handover? What does an associate do the week before an investment committee meeting? Then we work out which of those steps Copilot genuinely improves, and we train on those steps using their real documents.
Features come up along the way. They're just never the point.
This is also why adoption sticks after we leave. A feature is easy to forget. A workflow that saves somebody forty minutes on a task they do every single week doesn't get forgotten.
A simple way to explain it to your leadership
Buying Copilot licenses is like buying every tradesperson on site a new power tool. Then you run a session where somebody holds it up and shows that it spins. Impressive for about a minute.
Nobody uses it until you show them the specific cut they need to make, on the material they're actually working with, on the job they're already behind on.
Licenses are the tool. Utilization is whether it comes out of the box. Workflow is the cut.
What IT leaders should do next
- Pull your daily active license rate. From the Copilot dashboard, not monthly. Look at the real number before you do anything else.
- Ask ten users to name their task. Count how many can name a specific one. This takes an afternoon and tells you more than any report.
- Pick your three heaviest license roles. Write down the five things each of those roles actually does in a week.
- Find the overlap. Which of those fifteen tasks does Copilot measurably improve? It won't be all of them, and that's fine.
- Rebuild enablement around those tasks. Real documents, real users, in a room together.
- Set a baseline and remeasure in thirty days. Without a baseline you won't know whether anything changed, and you'll be having this same conversation at renewal.
Most organizations can finish the first two before lunch. The answers usually tell you everything you need to know about where you stand.
How Change Champions helps
We're a PROSCI-certified change management firm and a Microsoft Solutions Partner, and we've supported adoption for more than 60,000 users across financial services, manufacturing, maritime, utilities and heavy industry.
Our Copilot programs are built around workflows rather than features:
- Baseline and diagnostic. We measure where you actually are before anyone runs a session.
- Role-based workflow design. We map what your people do, then find where Copilot fits.
- Hands-on delivery. Prompt-a-Thons, instructor-led labs and office hours using your real documents.
- Champion networks. We build the internal community that keeps adoption going after we leave.
- Measurement that holds up. Weekly tracking against the baseline, reported in numbers your executive sponsor can defend.
Questions we get asked
What is a good Copilot adoption rate?
Measure daily active licenses, not monthly. Under 15% daily is struggling, 25% to 35% is typical, and above 60% is a well-run program. Monthly active figures flatter the result and hide the problem, because someone who used Copilot once on the 3rd counts as active all month.
Why aren't our Copilot licenses being used?
Almost always because people were taught what Copilot does rather than where it fits in their job. Feature training creates interest that fades within two weeks. Adoption sticks when the training is built around specific tasks people already do, using their real documents.
How long does it take to improve Copilot adoption?
Faster than most people expect once the approach changes. We've seen license utilization more than double within a month of a targeted program, and daily use move from a third of a team to nearly four fifths after a single well-designed workshop. The constraint is program design, not time.
How should we measure Copilot ROI?
Don't rely on self-reported time savings, because they consistently overstate the result. Measure daily active use, prompt volume per user, and completion time on specific named tasks you baselined before you started. If you didn't baseline, start now and treat the first thirty days as your reference point.
Do we need to roll out Copilot to everyone at once?
No, and it usually goes worse when you do. Start with the roles where you can name the workflows Copilot improves. Prove it there, build champions out of those users, then expand. A broad rollout with no workflow anchor is how organizations end up in the 95% that never leave the pilot.