Microsoft 365 Copilot costs roughly 30 EUR net per user per month, on an annual commitment. For 20 people that's about 7,000 EUR a year — enough to hire an intern for three months. Before you hand licenses out to the whole company, it's worth knowing where that money comes back and where it disappears without a trace.
We've been getting the same call in different variations for months now. "We bought Copilot because Microsoft pushed hard, but we don't see anything changing." We open the usage dashboard and find: 15 licenses, three people using it regularly, five who logged in once a month, the rest never. The invoice is right. The value is not.
Copilot isn't a bad product. But it isn't a "turn it on and it works" tool either. It only returns value in very specific roles, and only when the data underneath is in order.
What Copilot actually does
Set the conference demo aside. In daily work, Copilot does four things well: it summarises long email threads and Teams meetings, it pulls specific information out of your SharePoint documents when you ask, it generates first drafts of documents (proposals, reports, notes) in something close to your existing style, and it helps you write Excel formulas without knowing the syntax.
Useful — but each one has a condition. It will summarise your emails if your inbox isn't a 40,000-unread mess. It will find things in SharePoint if things are actually there — and if the person asking has permission to see them. It will draft an offer in your style if there are enough past offers to learn from. It will understand your Excel request if the sheet has proper headers, not "column1, column2".
Where we see a real return
In practice the return shows up in roles where a person produces a lot of text every day based on other text. Specifically: sales people preparing proposals from project history, project managers assembling weekly status reports, finance and controlling producing recurring reports, customer service replying to emails from a knowledge base, executives who get 30 emails a day and need to know what they're about without reading each one.
The common denominator: repetitive text work where the input already lives in Microsoft 365 (Outlook, SharePoint, Teams). When the input is in other systems — ERP, CRM, external drives — Copilot can't see it and won't help.
- A sales rep preparing 8-10 proposals a week — top candidate.
- A PM running 4-6 projects, stitching status reports — top candidate.
- Someone in controlling producing a monthly report from several sources — top candidate.
- A warehouse worker, a driver, a technician — no point. They don't work on text in Microsoft 365.
Where it's burned money
Three scenarios in which we almost never see Copilot pay off.
First: the data is chaos. SharePoint with no structure, everyone keeping files wherever they like, five versions of each. Ask Copilot "how much did we pay supplier X last quarter" and it returns something that sounds smart but is made up. People lose trust in a week and go back to old habits.
Second: the company has 8-10 people and everyone sits in one room. Meeting summaries are pointless — everyone was there. SharePoint search is pointless — you ask the colleague next to you. Here Copilot is competing against a solution that costs zero and has always worked: talking.
Third: licenses got handed out "equally to everyone" so nobody would feel left out. Result: 5% average usage, a decision not to renew after a year, and the conclusion "AI doesn't work in our company". Not true — 90% of those people were never the audience for this tool.
A real example
At one of our clients, a 40-person accounting and advisory firm, we ran a pilot on 8 licenses — not 40, deliberately. We picked three roles: two accountants running the largest entities, one tax advisor writing opinions, and five people from controlling who assemble monthly management reports for clients.
The setup took three weeks. We tidied the SharePoint structure (entities as libraries, consistent folder naming), captured a "your style" template for two document types, and set permissions so the tax advisor could see past opinions but not other people's entities.
After three months: the time to prepare a monthly management report for a client dropped from 6 hours to 1.5 (data still collected by hand, but the conclusions and narrative — with Copilot). The tax advisor prepares a first draft of an opinion in 30 minutes instead of 2 hours, then edits it himself. Annual license cost: about 2,900 EUR. Time recovered across those eight roles: about 55 hours a month, valued at those employees' rates — over 33,000 EUR a year.
It works. But if we had handed those 8 licenses out "one per department, fair share for everyone", the saving would have been zero and the cost the same.
How to pilot without regret
Four rules we stick to with clients.
Don't buy an annual commitment on 50 licenses out of the gate. Start with 5-10 for 3 months (Microsoft allows monthly billing if you negotiate through a partner). Pick roles, not the people who like gadgets.
Tidy SharePoint before you let Copilot loose in it. If Copilot surfaces an old password file or a purchase pricelist that everyone can see, you have a problem, not a saving. Copilot rides on SharePoint permissions.
Measure the process before the pilot. How long does it take today to prepare a proposal, a report, an opinion — actual minutes. Without a baseline, three months in you'll hear "well, faster I think" and won't know whether to continue.
Don't hand out licenses "as a reward". For many teams a better investment than Copilot is simply ChatGPT Team at 30 USD per user per month, without hooking it into company data. Cheaper, safe (if you set a policy), and enough for most use cases.
What next
Copilot is neither a miracle nor a scam. It's an expensive tool that pays back many times over in the right roles, and eats the budget in the wrong ones. The whole job is to answer honestly before you buy: who in this company produces a lot of text every day based on other text — and is that input already in Microsoft 365, or somewhere else.
If you're not sure whether Copilot makes sense in your business, talk to someone who runs these pilots before you sign an annual contract. The first call is free and takes half an hour.
Copilot pays off in specific roles on tidy data. "Same for everyone" is not a strategy — it's burned budget.