It is tempting to buy Copilot for everyone. It simplifies the rollout, it avoids awkward conversations about who gets access, and it is often presented as part of a larger deal. But a licence that is assigned and not used is still a cost, and Copilot is not cheap enough to treat as a rounding error.
For most organisations the better question is not whether everyone should have it, but how you decide who does, and how you will know whether it is working.
What the market looks like
Adoption is growing quickly. Microsoft’s proxy statement says Microsoft 365 Copilot passed 30 million paid seats during its fiscal year to June 2026, which one analysis puts at roughly one paid seat for every 15 commercial Microsoft 365 seats. On 28 October Microsoft added that the number of customers with more than 50,000 Copilot seats grew more than eightfold in a year.
So the market is not short of large deployments. But those figures count seats bought, not seats used, and they say nothing about whether each organisation got the value it expected.
Why blanket rollouts are risky
At a list price of 30 dollars per user per month, a Copilot licence costs 360 dollars a year. As an illustration, if a business assigns 1,000 licences and only 600 people use them regularly, the 400 idle licences cost about 144,000 dollars a year. That is an invented example, but the arithmetic scales to any size.
Idle licences are common across software generally. Zylo’s 2026 index found that organisations leave an average of 36 percent of SaaS licences unused. There is no reason to assume AI add-ons will be different, especially in the first months when enthusiasm and habit have not yet settled.
A better way to decide
Start with the roles most likely to benefit. These are usually people who spend a lot of time in documents, email, meetings and spreadsheets, and whose output can be checked easily. Give them licences, and tell them what you expect them to try.
Then track adoption over the first few months. Compare licences assigned with licences in active use, and look at which features are used. Microsoft provides usage reporting for Copilot in its admin tools, and it is worth agreeing in advance what level of use counts as healthy. If someone has not used it for a set period, such as 60 or 90 days, reassign the licence.
Review the numbers regularly instead of waiting for the next renewal. A monthly look is usually enough in the early stages, and it gives you evidence before you decide on a wider rollout.
Be clear about what you are buying
The commercial model is also changing. Copilot can be bought as seats, as usage-based credits or in a bundle. A Copilot Business promotion for organisations with 1 to 300 users, at 18 dollars rather than 21 dollars per user per month, runs to 31 December, while the larger enterprise promotions ended on 30 September. If you are considering E7, remember that it already includes Copilot, so the question becomes whether you want the whole bundle.
Ask how the price would change if you reduce seats, whether usage-based credits can replace some seats, and whether a promotional price carries into the next term. Ask also how AI add-ons are positioned in the wider deal, and whether a discount on one product depends on buying another.
A staged plan
A practical plan has four steps. Pilot with a defined group and a clear set of tasks. Measure assigned against active use after 60 to 90 days. Decide on expansion by role, based on what you saw. Review quarterly and reassign unused licences. Each step produces evidence that is useful both for internal approval and for negotiation.
Reading the adoption data
Raw usage numbers can mislead. A user who opened Copilot once in a month counts as active but may have gained little. A user who relies on it daily for one task may be extremely productive while showing modest numbers. Look at frequency and at which features are used, and combine the data with a short conversation with the people involved.
Be wary of averages. If a few enthusiasts use Copilot heavily and most do not use it at all, the average looks healthy while most of the licences are idle. Break the figures down by team and by role. That is where the decision about who should keep a licence is really made.
Common mistakes
The most common mistake is buying for everyone because it is simpler and then never looking again. The second is treating a promotional price as the real price, and finding at renewal that it has gone. The third is leaving ownership unclear, so that nobody is responsible for reassigning unused licences. The fourth is judging success by seats purchased instead of by the outcomes users report.
All four are avoidable. They need a small pilot, a defined review and a named owner, none of which is expensive.
What to tell the people who do not get one
A staged rollout creates a communication task. People who are not in the first group will ask when they can have access, and why not now. Give them a clear answer: licences are being assigned by role during a pilot, results will be reviewed at a set date, and wider access depends on what the pilot shows. A short, honest explanation prevents the pressure that often drives a blanket purchase.
Conclusion
Buying Copilot for everyone is simple, but it is rarely the best use of money. Start with the roles most likely to use it, track adoption closely and review the numbers regularly. Understand how the pricing model works and how it fits the rest of your Microsoft agreement.
How is your business deciding who gets AI tools? The organisations that can answer that with evidence will be in a far stronger position at their next renewal.
