There is a number doing the rounds in AI training sales pitches at the moment: “80% of Copilot seats go dormant within three weeks.” It gets quoted with a straight face, usually followed by a course offer. We went looking for the study behind it. We couldn’t find it.
Here is what the published data actually says. And it is worse.
The real numbers
Usage doesn’t collapse in three weeks. Adoption research shows it climbs for six to eight weeks, then stalls, typically with 15 to 25% of seats in active use. Across organisations that track it, only 20 to 30% of paid seats get used weekly. And in a Gartner survey of 123 IT leaders back in 2024, only four said Copilot was delivering significant value.
Why is “slower, then stalls” worse than the scary version? Because the made-up stat implies a quick fix. People gave up fast, so run a course and re-light the fire. The real pattern says adoption grows, then hits a ceiling and stays there. Ceilings are not an enthusiasm problem. They are a structure problem.
The gap nobody is paid to close
Walk round your business and ask who owns Copilot adoption.
IT? They own deployment. Licences assigned, security signed off, job done. L&D or HR? They own training. A lunch-and-learn ran in month one, job done. Managers? Nobody told them that visibly using the tools was part of the job. The owner? Assumed one of the above had it covered.
Everyone did their bit. Nobody owns the outcome. That is the ceiling. The keen 20% adopt on their own, and the other 80% wait for a reason that never arrives.
The stall has a price. Take 50 licences at $30 per user per month (Microsoft prices in dollars, even for UK businesses): 50 x $30 x 12 = $18,000 a year. At 20% active use, $14,400 of that is paying for shelf space. And that is before you count the improvement projects that never start because the team never got past “it writes emails”.
What owning adoption actually looks like
Four things. None of them technical.
- One named person accountable for adoption, with a number attached: cost saved and hours returned.
- Managers using the tools where the team can see them. Nothing kills adoption faster than a leader who pushes AI and never opens it.
- Training built on each team’s real work. Finance learns on month-end. Sales learns on proposals. Generic demos train nobody.
- Internal champions named early and given room to run small projects.
Notice that none of this is about the technology. Copilot adoption is not an IT project. It is a management decision. It costs nothing to decide who owns it. Most businesses never do.
The ceiling doesn’t wait for ready
Owners tell us two things: “we’re not sure we’re ready yet” and “it’s hard to get the team together”. Both fair. But the ceiling doesn’t care. Every month at 20% adoption is another month of paying full price for a fifth of the value. The firms pulling ahead are not the ones with more licences. They are the ones where somebody owns the outcome.
Our workshops exist for exactly this gap. We get the whole team in one room, working on your actual use-cases, and out the other side with a shared baseline and named champions. The CoPilot for Microsoft 365 workshop is 1.5 hours, in person, up to 25 seats, £1,295 (£1,295 / 25 = £51.80 a head). If you want to go further, the Interactive AI Half Day (3 hours, up to 25 seats, £2,495) sends teams away with working AI solutions and the groundwork laid for policy and champions.
One question to take to your next leadership meeting: “Who owns Copilot adoption here?” If the answer is “IT, probably”, you’ve found your ceiling.