Copilot adoption
Copilot adoption rarely fails on technology.
The licences work. The rollout is done. And still, usage stalls at summarising a meeting. Adoption is a behaviour problem, so you measure it as behaviour: before, during and 60 days after.
Why licences aren't a rollout
01 · The problem
Licences are a cost item until someone changes their behaviour. That's the whole case in one sentence, and it's the sentence most adoption plans skip over.
A typical programme looks like this: procurement arranges the licences, IT provisions them, there's a kick-off with a demo, and then everyone waits for usage to pick up by itself. After three months the counter shows eighty percent active users and it looks fine. Anyone who digs deeper discovers that nearly all that activity consists of a single action.
That's not surprising. Summarising meetings is the one task where Copilot explains itself: you don't have to learn anything, the button is right there, and the result is immediately usable. Every other task asks you to do something differently than you're used to, and that doesn't happen by itself, however good the tool is.
The second problem is that you can't demonstrate it. Without a baseline you don't know where people stood, so you can't show afterwards what changed. All you have is usage figures, and usage figures measure who pressed a button. For the conversation with your board that's too little, and rightly so.
That's why every programme here starts with five minutes of measuring. Role, work context, knowledge level and self-assessment, per employee. It costs the organisation nothing technically and produces the only figure you can use later: the difference between then and now.
What you get back
Per team you see the estimated time saved and the share genuinely applying Copilot. That backs up the investment and shows which team needs to go deeper.
Measure → Train → Embed
02 · The model
Step 1
Measure
A mandatory five-minute intake before the first module. Measures role, work context, knowledge level and self-assessment. Participants never see their own level: whoever reads 'beginner' starts behaving like one.
Step 2
Train
The training reorders itself based on the measurement: different sequence, different examples, different pace. Already know a module? Take the exemption test and skip it. The content is the same for everyone, the route isn't.
Step 3
Embed
It continues for 60 days after the certificate: spaced repetition of what went wrong, weekly usage checks and assignments that only unlock days later. Without this step it was just a nice afternoon.
The six mistakes we see most
03 · From practice
01
Starting without a baseline
The most common and the most expensive. You can't fix it retroactively: once the rollout is running, the starting point is gone. Everything you measure after that is a number without a comparison.
02
One session for a mixed group
A room with five levels becomes a compromise: too slow for the front-runners, too fast for the rest. First bring everyone to level with e-learning, then run a session on whatever is left.
03
Stopping at the certificate
A certificate measures that someone finished a course, not that their behaviour changed. Without repetition in the weeks after, it fades, and then the budget bought a feeling.
04
Putting up leaderboards
A leaderboard works for two weeks and then stops. What's left is people managing the number instead of their work. That's why our dashboard shows teams, not individual rankings.
05
Leaving the data a mess
Copilot is only as good as the findability of your documents. If three versions of the same policy coexist, it picks one, and not necessarily the valid one. That's an information problem, not a training problem.
06
Involving the works council too late
Any measurement involving employees raises questions about what the employer sees. Those questions come regardless; answering them early costs one conversation, answering them late costs a rollout.
Questions about Copilot adoption
04 · Frequently asked
What is Copilot adoption exactly?
The difference between handing out licences and getting people to work differently. So you don't measure adoption by active user count but by behaviour: are people doing tasks differently than six months ago, and are they keeping it up?
Why isn't active user count a good measure?
Because someone who summarises one meeting a day counts just as heavily as someone who rebuilt their entire working week. The number rises nicely and says nothing about value. It's also the number boards are quickest to settle for.
How long does a Copilot adoption programme take?
The training runs in short blocks over 60 days, at the participant's own pace. Reinforcement continues after the certificate. For an organisation of a few hundred people, count on three to four months between baseline and follow-up measurement.
Do you need access to our Microsoft environment?
No. We measure through the participant, so you can start tomorrow without an IT project. No fuss, just safe; the details are in our privacy statement.
What usually goes wrong?
Three things, in this order. The rollout starts without a baseline, so nobody can demonstrate afterwards what it delivered. One classroom session is given to a group spread across five levels. And after the certificate nothing happens, so it fades within two months.
Should everyone start at once, or in phases?
In phases, but not by department, by readiness. Start with the teams whose manager uses it themselves. That group produces the examples you'll use to convince the next one. A rollout starting with the most sceptical department costs twice the energy.
Read on
- The plan in depth
Measure, train and embed, written out step by step.
- The employer dashboard
The dashboard, the monthly email and the quarterly report.
- AI literacy and the obligation
Since August 2026 the same measurement doubles as your AI Act justification.
- Copilot for HR and L&D
Figures you can use to justify the rollout internally.
Start by measuring. The rest follows.
In a half-hour demo we show the baseline assessment and the dashboard on sample data. You get an answer from a person within two working days.