The seats get switched on, someone runs a one-hour demo of what Copilot can do, and three months later the usage report shows a handful of people using it every day while everyone else tried it once. Nothing went wrong in a way anyone could point to. Although people were shown the tool, they were never shown their own work done with it, which is the only thing that makes a habit.
What follows is the first month as we run it. It is our method rather than Microsoft's, and it assumes you are training a first team of people.
Why a demo is not training
A demo shows features: sum this up, draft that, find the other. The people watching nod, because every feature makes sense, and then go back to their desks with no idea which of the dozens of things they do this week it should touch. Training starts from the other end. From the work each person does, it picks out the three or four pieces where Copilot saves them time without creating a checking job bigger than the saving.
Microsoft's own Copilot Academy, inside Viva Learning in Teams, is a good course on the features and worth pointing people to. It cannot know your rules, your clients or which of your people's tasks are worth changing, and that is the part that decides whether the seats get used.
Before day one
Three things need to be in place before anyone is trained, because training on top of their absence teaches the wrong habits.
- The rules. People need to know what may go into Copilot and which output needs a person to read it before it leaves. If that is not written down yet, what an AI policy should contain sets out what to settle.
- The permissions. Because Copilot answers from whatever each person can open, check what the team can reach first. What Copilot can already see has four checks.
- The before. Each person writes down how long their chosen tasks take now and who checks them. Without that, nobody can say at the end of the month whether anything changed, which is the most common reason pilots fade.
Week one: three tasks from their own work
Each person picks three tasks they do every week, and the session is spent doing those tasks with Copilot on their own material while someone sits beside them. For most people the first three are some mix of catching up on a missed meeting, drafting a reply to a long email thread, turning notes into a first draft and finding something they know exists but cannot locate.
The one skill worth teaching explicitly is how to ask. A request that says who the output is for, what a good result looks like and what to leave out gets a usable first draft far more often than a one-line instruction. When it is their own task on the screen, people pick that up in an afternoon.
Week two: a daily habit and a place to ask
The aim for week two is that each person uses Copilot on their chosen tasks every day, busy days included, because a tool used only when there is time never becomes the way the work gets done.
Give the team one place to ask questions and share what worked, such as a Teams channel, and have someone answer within the day. Collect the requests that produced good results into a shared list, written in your business's words. Once the month is over, that list becomes the starting kit for the next team.
Week three: where it goes wrong
By the third week people trust the output more than they did on day one, which is the right moment to show them where it fails. Copilot writes fluently whether it is right or wrong. It can present a figure nobody checked, attribute a remark to the wrong person in a meeting summary or draw on an old version of a document.
Teach one more thing in this week. If Copilot shows someone a document they should not be able to see, such as a salary file or a board paper, that is a permissions finding. People need to know it gets reported, who it goes to and that reporting it will never get them into trouble.
Week four: did it work?
Go back to the before. For each person's chosen tasks, compare how long they take now and how much checking the output needed. The people who do the work judge the result, because they know whether a saving held up or just moved the work somewhere else.
The usage report in the Microsoft 365 admin center shows who used Copilot, in which applications and when they last did. Names are hidden by default until an administrator turns them on. Together with the before and after, it tells you who keeps a seat, what the next team should be taught first and whether the business is ready to widen.
The managers in the room
Managers need a short session of their own. They decide which of their team's work Copilot may touch, they are the reviewers for much of the output and they set the tone on whether using it is encouraged or frowned on. When a manager never mentions Copilot, the team treats it as optional and stops using it.
Where this fits
This month is Adopt, the second phase of Track, which ends with a test on the work your team does. After it, the improving part of Trackside keeps the shared list current and fixes whatever people keep rejecting. If you are still weighing the cost, what Microsoft 365 Copilot costs sets out the parts around the seat, including this one. We are a Microsoft Solutions Partner for Microsoft 365 Copilot.