You've deployed Microsoft 365 Copilot. Licences are assigned. Now the awkward question arrives from leadership: is anyone actually using it? This guide explains how to measure adoption, what the data means, and how to close the gap between licences purchased and value delivered.
This is not a readiness guide. If you haven't deployed Copilot yet, start with our Copilot Readiness Checklist first, then come back here once it's live.
Microsoft 365 Copilot licences cost approximately £25–30 per user per month. For an organisation with 100 licensed users, that's £30,000+ per year. If only 30 people are actively using it, you're burning £21,000 annually on unused AI capacity.
The uncomfortable reality is that Microsoft's own research shows adoption curves for Copilot are slow without active management. Initial enthusiasm from early adopters drops off within 6–8 weeks unless use cases are embedded into daily workflows.
Microsoft provides Copilot usage data in three places, each with different depth and access requirements:
Navigate to Reports → Usage → Microsoft 365 Copilot. This gives you a 7, 30, or 90-day view of active users per workload (Teams, Word, Excel, PowerPoint, Outlook, OneNote, Loop). You'll see enabled users vs active users — the gap between these numbers is your adoption problem.
The limitation: it's a static snapshot with no trend line and no per-user breakdown unless you enable detailed reports.
Viva Insights provides per-department adoption breakdowns, prompt quality signals, and time-saved estimates. Requires Viva Insights licences and takes 2–4 weeks of data to produce statistically meaningful output. Best for organisations past the 60-day mark.
The /beta/copilot/reports/getMicrosoft365CopilotUsageUserDetail Graph endpoint returns per-user Copilot activity across all workloads for the last 7, 30, or 90 days. This is what M365Clarity uses in the Copilot & AI tab. It requires Reports.Read.All permissions and returns aggregate-only data by default (privacy settings permitting).
48-hour lag: Graph activity reports for Copilot have a 48-hour data lag. If you've just run a scan and see no data, wait two days and re-scan.
Active users ÷ licensed users × 100. This is your headline number. Microsoft defines "active" as at least one Copilot interaction in the period. Aim for 70%+ by 90 days post-deployment. Below 40% at 90 days signals a structural enablement problem.
Which apps are users actually using Copilot in? Teams meeting summaries tend to be the fastest-adopted feature (lowest friction — it just works passively). Word and Outlook drafting come next. Excel and PowerPoint Copilot typically lag because they require more intentional use.
| Workload | Typical early adoption | Barrier to adoption |
|---|---|---|
| Teams (meeting summaries) | High | Transcript must be enabled |
| Outlook (email drafting) | Medium | Prompt quality learning curve |
| Word (document generation) | Medium | Requires use case awareness |
| PowerPoint (slide creation) | Low | Brand templates not yet integrated |
| Excel (data analysis) | Low | Requires clean table-structured data |
| Loop (collaborative notes) | Very low | Loop itself has low adoption |
Not just whether users opened Copilot, but how much they used it. Low prompt counts (1–5/week) suggest users are experimenting but haven't found a core workflow to embed it in. High prompt counts (20+/week) indicate power users — these are your internal champions to identify and amplify.
Are users who tried Copilot in week 1 still using it in week 4? This is the most honest measure of whether adoption is real or just novelty-driven. You need 90-day data to track this properly.
Are the right people licensed? Sales teams generating proposals benefit more from Word and Outlook Copilot than, say, warehouse staff. If your highest-usage department has 20 licences and your lowest-usage has 50, that's an immediate reallocation opportunity.
Copilot relies on a well-configured tenant. If meeting transcriptions are off, Teams Copilot can't summarise meetings. If sensitivity labels aren't applied to documents, Copilot will surface data it shouldn't in summarisation. Run a configuration scan before blaming users.
The configuration prerequisites that most commonly block Copilot value:
M365Clarity checks all of these. The Copilot & AI tab in M365Clarity shows your Copilot licence tier, active adoption by workload, and the seven governance controls that determine whether Copilot is operating safely. Run a free scan.
Generic training ("here's what Copilot can do") doesn't drive adoption. Specific use cases embedded into existing workflows do. The highest-ROI use cases by role:
Many users open Copilot, stare at the prompt box, and close it. They don't know what to ask. The fix is a prompt library — a shared document (ironically, best created with Copilot) containing 20–30 tested prompts relevant to your organisation's workflows. Pin it to relevant Teams channels.
Users worry about accuracy. Copilot makes mistakes — it hallucinates occasionally, summarises meetings imperfectly, and misinterprets context. The solution isn't to hide this; it's to position Copilot as a first draft tool, not a finished output tool. "Copilot gets you 80% of the way there in 20% of the time" is more honest and more useful than overpromising.
For most organisations, a simple monthly adoption review is sufficient. What to track:
Present this monthly to IT leadership and quarterly to the business. Frame it in cost terms: "We have 100 licences. 68 users are active. Our effective cost per active user is £X. Last month it was £Y. Here's the plan to close the gap."
Usage data tells you adoption. ROI requires you to measure outcomes. Microsoft's own Copilot Impact Report (2026) cites:
These are averages — your organisation's numbers will vary significantly. The most credible way to measure ROI is to run a structured 90-day pilot: take two matched groups, give Copilot to one, run identical workflows, and measure time-to-completion on specific tasks. This gives you a defensible ROI number grounded in your actual work rather than industry benchmarks.
Adoption without governance creates risk. As Copilot usage grows, so does the surface area of your data estate that AI can touch. The governance controls that matter most:
M365Clarity's Copilot & AI tab shows your licence tier, active users by workload, 30-day adoption trends, and the governance controls that determine whether Copilot is operating safely. Free to scan one tenant.
Run a free scan →Measuring Copilot adoption is straightforward with the right data. The harder problem is acting on it. Most organisations see three distinct cohorts: early adopters who've found real use cases (protect and amplify these), casual experimenters who tried it but haven't embedded it (focus enablement here), and non-users who haven't engaged at all (investigate whether licences are correctly assigned).
The organisations getting real ROI from Copilot are treating adoption as a change management problem, not a technology problem. The technology works. The question is whether you've given people a reason to change how they work.
Related articles