Adoption

You've bought Copilot licences. Here is where the value gets lost

By Matteo Garoli 6 min read

A white outline icon of three people standing together on a frosted glass tile, with warm amber light and one ice blue orb glowing behind it on a near-black background.

A Microsoft 365 Copilot licence gives a person access to a tool. It doesn’t change what that person does on a Tuesday morning. Microsoft’s own adoption score counts days of actual use. Microsoft’s documentation now calls the product Microsoft Copilot.

I work as a Microsoft 365 Copilot Cloud Solution Architect. This year I’ve run Copilot adoption reviews with large organisations in the UK and across Europe, and I teach people to use AI in their daily work. My view is that adoption is a change of habits. It starts with training people on their own tasks and carries on long after the launch email.

Is buying Copilot licences the same as adopting Copilot?

No. A licence is access. Adoption is people using Copilot, repeatedly, as part of their normal work.

Microsoft describes the AI adoption score in Adoption Score, in the Microsoft 365 admin center, as the extent to which licensed users use Copilot as a daily habit. A score of 100 means licensed users average at least three days of use a week, which is 12 of the past 28 days. Microsoft says people who reach that threshold in a month are highly likely to become long-term users. It counts intentional actions in apps such as Outlook, Teams, Word and Excel, so a licensed person with no active days scores zero.

A pilot at the UK Department for Business and Trade shows what that looks like. The department ran 1,000 licences for three months in late 2024, about 70% of them given to volunteers, and published an evaluation in August 2025. According to the Microsoft dashboard data in that report, around 64% of licence holders used Copilot at least once a week and around 30% used it on an average working day, with 1.14 actions per licence per working day. It was one department and one pilot, so I treat it as a single data point. Even with mostly volunteers, it shows a shape worth planning for: more people try Copilot than build a habit with it.

What goes wrong after the licences are bought?

The gaps I look for first are these: nobody has chosen the tasks Copilot is for, the training is one demo for everyone, there are no champions, and nothing happens after the launch email.

Without chosen tasks, a licence gives people a blank prompt box. Microsoft’s adoption guidance starts from team goals, asks for high-impact use cases that ladder up to them, and recommends a first cohort working on high-value, low-risk scenarios.

A single generic demo covers nobody’s work well. In the Department for Business and Trade evaluation, interviewees were generally positive about the formal sessions but described the departmental and Microsoft ones as often introductory or high level, and less relevant to their roles. They found their own hands-on practice more useful because it showed them where Copilot helped in their job. The report links at least two hours of independent training to higher satisfaction, while for departmental sessions it found a measurable difference only at five hours or more.

Without champions, nobody close to the work shows it working. Microsoft’s planning checklist says to pick champions and early adopters from people who share a process or department and are willing to start first. The same evaluation found that the attitude of a person’s team and line manager appeared to influence how much and how they used Copilot. At one large organisation I reviewed, the strategy, the team and the governance were solid, but the champions were informal and nobody recognised their work. At one public body the real job was moving adoption out of the central team and into the individual departments.

The IT admin onboarding guide on Microsoft Learn has five steps, and the last one is “Send a welcome email and enable feedback”. For adoption material, the guide sends you to the Microsoft Adoption site, which describes what comes after launch: a community of practice, user feedback surveys and periodic service health reviews built on Microsoft 365 admin center and Copilot Dashboard data. Microsoft calls those reviews critical to the long-term success of the transformation. A programme that ends at the welcome email has none of them.

How should you train people to use Copilot?

On their own tasks, team by team, in small sessions, and more than once.

I start with the work a team already has, for example a weekly report or a supplier email thread. Each session runs on the team’s own files, and everyone leaves with something they can repeat on Monday. The Department for Business and Trade found self-led practice beat departmental sessions on satisfaction, so my sessions are mostly hands-on practice on the team’s own files, with little time spent presenting. I begin with simple uses that give a result in the first week and move to harder ones once the habit exists.

Microsoft’s Interactive Scenario Library, with example outcomes by industry and role, works as a menu when a team can’t say where to begin.

I have made the one-size-fits-all mistake myself. My first follow-up proposal for one customer was built on another customer’s and included a service they had just told me was not a priority. I rewrote it from what they had said.

For companies I propose training by department, because legal and operations don’t need the same session. I plan half days of four hours and a refresher about every ninety days, since in my experience a session ages in about three months, and I want someone following each team in between.

How do you measure Copilot adoption?

By habits, not by licences assigned. Microsoft’s own yardstick is the average number of days a licensed person is active, with a target of three a week.

Usage tells you where a habit is forming. It doesn’t say why one isn’t forming elsewhere, so I talk to the people in the teams that fall short. It also says nothing about whether the work got better: the Department for Business and Trade found small time savings, largest on written tasks, and no evidence that they had improved productivity. I cover which numbers to read in a separate article on measuring Copilot adoption.

In what order should you fix a stalled Copilot rollout?

Tasks first, then training on those tasks, then champions, then measurement, then a rhythm that continues after launch. Before any of it, check what Copilot can reach: Microsoft’s plan stage includes securing the data foundation, and I cover the permissions side in a separate article.

  1. Choose two or three tasks per team and one first cohort, all from work the team does every week.
  2. Train that cohort on those tasks, team by team, on their own files.
  3. Turn the people who make it work into champions in the same teams, and tell their managers, since team and line manager attitudes shaped use in the Department for Business and Trade pilot.
  4. Track average days of use each week for ninety days and talk to the teams that fall short of three.
  5. Repeat the training about every ninety days and review the results with the business owners. Widen the rollout or right-size licences after that; right-sizing is part of Microsoft’s manage stage.

If your Copilot rollout has stalled and you want a second opinion on where to start, write to me.

Sources

If you are working on Copilot adoption, a training or a talk, or you are starting out in this world, write to me.