Module 7 of StartCloud's Copilot Essentials learning pathway, in five short units with a knowledge check: working out your break-even in hours per seat, which Copilot reporting numbers are real signal and which are vanity metrics, how to run an eight-week trial with a control group on the included Copilot Chat, the honest signs Copilot is not working for a role, and how to reassign or cut seats well.
Is Copilot Actually Worth the Licence?
What "worth it" actually means
Somewhere in your business there is a Copilot licence quietly renewing every month, and nobody can say with a straight face whether it is earning its keep. That is not a failure of nerve. It is that nobody agreed what "working" would look like before the licences were switched on, so there is nothing to compare against now.
The good news is that the maths for a small business is refreshingly simple. A seat is worth it when the time it gives back is worth more than the seat costs. Work out your break-even in hours, once, and write it down. Take the monthly cost of a seat, divide it by what an hour of that person's time is worth to the business, and you have the number of hours a month Copilot needs to save to pay for itself. For most office roles it lands somewhere between one and three hours a month, which is a much lower bar than the sales deck suggests.
- How many hours a month does this seat need to give back to break even?
- Which specific tasks is it giving those hours back on, named out loud?
- Would this person notice, and complain, if the licence disappeared next Friday?
If you cannot answer the second question, the first one does not matter yet. "It feels useful" is a lovely sentiment and a terrible basis for a renewal.
Copilot is not one investment, it is a handful of small separate ones. It can be brilliant for the person writing proposals all week and close to useless for the person next to them, and both of those results are correct at the same time. Judging it as a single company-wide yes or no is the fastest way to get the decision wrong in both directions.
Microsoft gives you real reporting for free. The Microsoft 365 admin centre has Copilot readiness and usage reports, and the Copilot Dashboard in Viva Insights adds adoption, impact and sentiment. Module 5 of this pathway walks through where they live.
The trap is that the friendliest number in those reports is the least useful one. The usage report counts someone as an active user if they tried a user-initiated Copilot feature at least once in the period you selected, and that period can be 7, 28, 90 or 180 days. One prompt in six months turns someone green. That is a licence with a pulse, not a licence doing work.
- Active days per user, which the admin centre reports show alongside the headline numbers
- Whether usage holds up in weeks five to eight, long after the novelty
- Named tasks that have genuinely shrunk, in each person's own words
- Adoption by app, so you can see whether it is being used in real work or only in chat
- Self-reported time saved, collected the same way every fortnight
- Licences purchased, which measures your purchasing, not your value
- "Everyone tried it in week one", which measures novelty
- Total prompts submitted, which rewards chatting, not finishing work
- The active user rate on its own, since one action in the whole period counts
- How enthusiastic the person who championed the purchase still is
The number most small businesses actually need
Self-reported time saved gets sniffed at, and it should not be. If you ask the same five people the same question every fortnight, in the same words, you get a trend line that is good enough to make a decision with. Something like: "roughly how much time did Copilot save you this fortnight, and on what?" Two minutes each, recorded in a spreadsheet. It will not survive an audit, but you are not writing a journal paper, you are deciding whether to renew five seats.
The Microsoft 365 Copilot impact report compares how Copilot users and non-Copilot users work across groups, and shows Copilot assisted hours. There is also a business impact report where you can bring your own organisational metrics in. It takes some setting up and it is far more than most small businesses need, but if you already have the licensing, the comparison is done for you.
Microsoft's own guidance is to pilot, then deploy, then operate. That is sensible, and it is missing the one ingredient that turns a pilot into evidence: something to compare the pilot against. Without it you are measuring enthusiasm, and enthusiasm is exactly what fades in week five.
A trial with a control group sounds like something a large company does. It is not. In a fifteen person business it means five people with licences, two people without, eight weeks, and one date in the calendar.
Write down what you are trying to shrink
Pick two or three real tasks per person and get a rough baseline for each. Not precise, just honest.
- "Monthly report commentary: about half a day"
- "Meeting notes: 20 minutes after every client call"
- "First draft of a proposal: two hours"
Keep a control group on Copilot Chat
Pick a couple of people doing similar work who stay on the included Copilot Chat for the length of the trial. They are your comparison, and they are not being punished.
- Same sorts of tasks, same sort of week
- They still get AI, just without the work data grounding
- Ask them the same fortnightly question as everyone else
Eight weeks, with a decision date in the calendar
Long enough to get past the novelty, short enough that nobody forgets it is a trial. Book the decision meeting on day one so it cannot quietly slide.
- Weeks 1 to 2: training and the fun phase, do not measure anything yet
- Weeks 3 to 8: the honest stretch, measure fortnightly
- Decision day: keep, reassign, or stop, no fourth option
Ask the same three questions, in the same words
Consistency beats sophistication. Changing the question halfway through ruins the trend line, and the trend line is the whole point.
- Roughly how much time did it save you, and on what?
- What did you try that did not work?
- Would you hand the licence back if we asked?
If the licensed group got prompt training and the control group did not, you have measured training, not Copilot. Give everyone the same grounding in writing prompts that work before the trial starts. It costs an hour and it is the difference between a result you can act on and a result you can argue with forever.
Here is the part that gets skipped in every vendor conversation. Sometimes the answer is no, and the business that says no on time is in far better shape than the one paying for shelfware because cancelling would feel like admitting a mistake.
There is nothing embarrassing about a seat that did not land. You ran a fair trial, you got a clear answer, and you stopped spending on something that was not helping. In most businesses that is called good management. Sunk cost is not a strategy, and neither is hoping the next quarter will be different.
The honest signs it is not working for a role
Highly specialised writing, or a role that lives in systems outside Microsoft 365. Copilot cannot ground itself in a system it cannot see, and no amount of training changes that.
In roles where every number and clause has to be verified anyway, the review time can wipe out the drafting time. That is a real result, not a training gap.
A spike after every nudge and a flat line in between is the clearest signal there is. People use tools that help them without being asked.
Eight weeks in, if the answer is still "it is handy for the odd thing", the honest read is that the seat has not found its job.
- 1Reassign before you cancel. Someone on the waiting list may be a far better fit, and the licence is per user.
- 2Say plainly that the tool did not suit the role. Nobody failed, and framing it as failure teaches people to fake usage next time.
- 3Write down what you learned about which work Copilot suits in your business. That note is worth more than the trial.
- 4Keep those people on Copilot Chat, which is included anyway, so they still have a sanctioned AI tool.
- 5Set a date to revisit, perhaps six months out. Copilot changes constantly, and so does the work.
In the Perth businesses we work with, the pattern is rarely all or nothing. It is usually a handful of seats that are clearly worth keeping, one or two that get reassigned to someone who was quietly desperate for them, and a couple that get handed back without drama. That is not a failed rollout. That is a business that now knows exactly where AI helps and where it does not, which is more than most of your competitors can say.
- Microsoft Learn: Microsoft 365 Copilot usage report
- Microsoft Learn: Microsoft 365 Copilot readiness report
- Microsoft Learn: Copilot Analytics introduction
- Microsoft Learn: Copilot Control System measurement and reporting
- Microsoft Learn: Microsoft 365 Copilot impact report
- Microsoft Learn: Set up Microsoft 365 Copilot and assign licenses
Details were current at the time of writing (August 2026). Microsoft renames and reshuffles the Copilot reports regularly, so the screens may not match exactly by the time you go looking.
Knowledge check
5 quick questions. Get 4 right and the module is yours.
1. Why is the active user rate a weak measure of whether Copilot is working?
2. What makes a Copilot trial produce evidence rather than enthusiasm?
3. Which of these is a fair way for a small business to measure time saved?
4. Eight weeks in, one person's usage only spikes the day after you remind them. What does that tell you?
5. You decide to hand back two of five Copilot seats. How should that be framed?