You bought Copilot licences. Why are people only using them for email?

Because buying the licences is the easy part, and almost nobody redesigns the work afterwards. People open Copilot, use it to summarise an email or tidy a paragraph, then go back to working as before. The tool works. The trouble is that it gets treated as software to install, when it is a change to manage.

What does the data on Copilot adoption in companies really say?

Almost every large company has Copilot, and few have taken it to scale. The two figures have to be read together.

Microsoft says more than 90% of Fortune 500 companies use Microsoft 365 Copilot (earnings call of 29 October 2025). It is the vendor’s own figure, and it measures use in the loosest sense: a handful of licences is enough to be counted.

Forrester tells the other half. In its analysis The Copilot Reality Check, dated 27 February 2026 and written after dozens of conversations with CIOs and CDOs, it describes most enterprises as still in pilot mode, with demand it calls “disciplined, governed, and conditional”. In the analyst’s view, most are still 12-18 months from a rollout at scale.

Meanwhile the bill keeps growing (we break it down line by line in how much Copilot really costs). The list price of Microsoft 365 Copilot for large enterprises is $30 per user per month on an annual subscription (€26 in Italy, excluding VAT), and it comes on top of the Microsoft 365 licence the company already pays for. For a hundred people that is $36,000 a year (€31,200 at the Italian list price), enough for someone to ask, sooner or later, what it is producing.

We found no public Italian data specific to Copilot, but the wider picture points the same way. The HR Innovation Practice Observatory at Politecnico di Milano, in figures presented in May 2026 and picked up by Fullpress, finds that 44% of Italian workers use artificial intelligence, mostly as an assistant for repetitive tasks and not as a lever for changing processes. Only 9% of organisations manage the time AI saves in a structured way.

In conversations with clients we see the same pattern: the licences are there, and real use concentrates on two or three basic functions. The question we hear most often from the people running IT is always the same: “why are we paying for it, if they only use it for email?”

Why does a pilot that “goes well” so often lead to no rollout?

Because it measures the wrong thing. A pilot with no serious errors, lukewarm positive comments and nobody complaining looks like a success. That does not mean it generated adoption.

Most Copilot pilots we see start by counting “how many people have logged in” instead of asking “which task, manual before, now runs inside Copilot in a repeatable way”. The first number always goes up, because a login is enough. The second stays at zero until somebody removes a step from the existing process and hands it to the assistant, and while it sits at zero, positive comments count for little. How you get from a pilot like that to a real rollout is in from pilot to rollout.

If it is not a technology problem, what is it?

A process problem. Copilot inside Word, Excel or Teams does what it promises for people who already know what to ask it and at which point in their work to use it. Almost no organisation did that redesign work before handing out licences: it assigned them, often to a whole department, and let everyone work it out alone.

In our experience the difference shows when you compare an “installed” use case with a “built” one:

	Installed (the most common approach)	Built (the approach that generates adoption)
Starting point	Licence assigned to a department	One specific task, manual today, chosen as the target
Who decides how it is used	Each person, alone	The process is redesigned to include Copilot at a precise point
What gets measured	Logins, “active users”	Time saved on that task, repeated over time
Training	One generic webinar, once	Side-by-side support on the real use case, until it becomes habit
Who answers for the result	Nobody in particular	One person who answers if adoption does not rise
Typical outcome after six months	Use shrunk to emails and summaries	The chosen task runs steadily inside Copilot

That is why every project of ours starts with a map of the work. It is the first of the three stages we work in: Map, Build and Hand over. First we understand where the work stands today and what it costs to do it that way. Then we build inside the tools the company already uses and redesign the piece of process Copilot can genuinely replace. At the end the use case passes to your team, which becomes responsible for it. It is the work we do in our adoption and change management projects.

A concrete example: at a consulting firm with more than a hundred consultants we took a single task, preparing presentations and proposals, and built it inside Copilot. A presentation went from four hours to under one, and more than 115 people use it today.

How do you measure whether people with a Copilot licence really use it?

By looking at who uses Copilot repeatedly on a specific task (we collected the public figures on real use in how many employees really use Copilot). The login count, which most dashboards show first, only tells you who opened the tool.

Even the soundest public figures count who has access or who opens it: none of those we checked says how much of a team’s daily work has moved inside Copilot. So the measure has to be built in house. You choose the task before the pilot, measure how it looks today and compare afterwards, with the same people.

There is already a free starting point: Microsoft’s Copilot Dashboard shows readiness, adoption and impact, and for Microsoft 365 business and enterprise customers it needs no additional licences. It tells you who uses Copilot and how much; which task has moved inside it is for you to establish.

That task-level number is also the only one worth taking to whoever signed off the licence purchase, and the starting point for measuring the return on Copilot.