Copilot Licences Turned Into Daily Use at a 1,400-Person Professional Services Firm
The firm had bought 900 Microsoft 365 Copilot licences and renewal was eleven months away. Most holders had tried it once. We rebuilt enablement around role-specific scenarios and took the share qualifying as power or habitual users in the Copilot Dashboard from 22% to 71%.
About the Client
The Challenge
The firm had committed to 900 annual Microsoft 365 Copilot licences on the strength of a demo and a board expectation that it would be seen to be using AI. Microsoft had reported over 30 million paid Copilot seats in its FY26 Q4 results and the board had read the headline. Nine months on, the only usage data was an anecdote: it summarised email threads, and nobody could name a second thing it did.
The actual numbers had a familiar shape. In the Copilot Dashboard's classification nearly everybody sat in the novice band: at least one action in the last twelve weeks, activity in fewer than nine of them. Usage concentrated in Teams meeting recaps and Outlook thread summaries, the two features that require no change in behaviour because they appear where the user was already looking. Word, Excel and PowerPoint were close to nothing, and Copilot Chat grounded on work data was barely touched.
The rollout had been run by IT, and run well by IT standards: clean assignment, a launch email, a recorded webinar, a page listing features. None of that changes how a tax manager writes a client memo. The webinar showed what Copilot can do; it never showed one role the task they do every Tuesday.
There was a second problem underneath. Where people tried substantial prompts — what does the firm know about this client — the answers came back thin, because the material sat in badly organised SharePoint sites, and a weak answer on the first serious attempt makes a permanent non-user.
Our Approach
We treated adoption as a measurable programme with a baseline, a cohort structure and a weekly number, not as a launch event followed by hope.
Baseline and Usage Forensics (Weeks 1-2)
Stood up the Copilot Dashboard in Viva Insights properly — it needs the app deployed and a reporting population defined before the numbers mean anything. Fixed the vocabulary first. An active user is anyone with one Copilot action in the previous 28 days, a very low bar. A power user runs 15 or more actions a week in at least nine of the last twelve weeks; a habitual user clears the same nine-week test at one to fourteen. Those two describe a habit, the active count a login. Baselined all four by practice group, allowing for the reporting delay of up to six days.
Scenario Design by Role (Weeks 2-5)
Built enablement around named tasks rather than product features. For each of 6 job families we wrote scenarios as a real deliverable: interview notes turned into a findings document, a proposal assembled from three prior bids, a client meeting prepared from six months of correspondence. Each carried a tested prompt, the expected output and a note on where Copilot gets it wrong. 18 were published.
Content Readiness for the Scenarios (Weeks 3-7)
Every scenario was tested against the real tenant before publication. Where one failed because the source material was unfindable we fixed the content rather than dropping the scenario — consolidating stray sites, adding metadata, clearing sharing links that surfaced material people should not see. Findings fed the parallel Copilot readiness workstream.
Champions and Cohort Rollout (Weeks 5-10)
Recruited 34 champions — one per practice group, chosen for credibility with peers rather than technical interest — and gave them the scenarios two weeks early. Ran enablement in small cohorts: a 45-minute session, a two-week practice window with one assigned scenario, then a clinic. That window is what most programmes leave out and what produces the habit.
Measurement, Value and Licence Governance (Weeks 9-14)
Reported the power and habitual bands by group weekly and treated a flat cohort as a defect. Set expectations on the dashboard's Impact figures before the finance director saw them: Copilot assisted value is calculated from a default of $72 an hour, which is a configurable assumption rather than a measurement of this firm's cost base. We substituted the firm's blended rate and labelled which part of the number was modelled. Switched on external benchmarking, which returns nothing until twenty companies sit in the comparison group. A licence unused for sixty days is now reclaimed for a waiting list, which moved 180. Rebuilt renewal modelling against the 1 July 2026 pricing: Copilot bundled into Business Standard with Copilot at $23.50 and Business Premium with Copilot at $32.00, suite prices raised the same day, and Microsoft 365 E7 added — E5 plus Agent 365, Microsoft 365 Copilot and the Entra Suite.
Solution Architecture
Measurement: Copilot Dashboard in Viva Insights reporting the power, habitual, novice and other bands per app, with groups scoped above the privacy floor that suppresses small cohorts
Scenario library: Role-indexed scenarios with tested prompts, sample outputs and known limitations, surfaced in the Copilot Prompt Gallery
Grounding: Microsoft Graph and the semantic index across SharePoint, OneDrive, Exchange and Teams, remediated site by site through SharePoint content readiness
Enablement operating model: Champion per practice group, cohort waves, a two-week practice window and a monthly clinic that outlives the programme
Licence governance: Sixty-day inactivity reclaim with a waiting list, reviewed monthly against the renewal date and July 2026 bundle pricing
Results
Client identity withheld under confidentiality. Figures are representative of engagements of this scope.
Technologies Used
Key Takeaways
If your organization is facing a similar challenge, here's what we learned:
Licences assigned is not a metric, and neither is active use. Active means one action in 28 days. Report the power band instead — 15 or more actions a week in at least nine of the last twelve weeks — with the habitual band beside it. Those thresholds separate a habit from a trial, and they are already computed.
IT-led rollouts underperform business-led ones for a structural reason. IT can assign the licence and answer the tickets. IT cannot tell a bid manager which part of writing a bid to hand over. The people who can sit inside the practice, which makes the champion network the delivery mechanism.
The first serious prompt decides the user. Someone who asks a real question and gets a thin answer does not try a third time. Content readiness is an adoption dependency, not a separate project. A feature tour produces recognition; a scenario written as a deliverable produces a habit.
There is no credible external adoption percentage to compare yourself against. The figures circulating in vendor articles trace back to each other rather than to Microsoft. The only defensible comparison is the Copilot Dashboard's external benchmark, which returns nothing until twenty companies sit in the group. Treat assisted value the same way: the $72 hourly rate is a default assumption, not your cost base.
Build the reclaim policy before renewal, not during it. A sixty-day reclaim run monthly gives you a defensible renewal number, and the July 2026 bundle and suite pricing changed what the renewal options are. We place Copilot specialists into adoption teams.
Related Case Studies
Are Your Copilot Licences Being Used?
Most firms cannot answer with a number. We baseline the power and habitual bands, build scenarios per role, and get adoption to a level that survives renewal.