Copilot licensing follows a pattern so common it's nearly a law: bought in enthusiasm, assigned in bulk, reviewed never. A leadership push or a pilot program puts seats in everyone's hands; a year later the subscription renews on autopilot, and nobody in the building can say what those seats produce. At roughly $30 per user per month, a 300-seat deployment is a $108,000 annual line item — which makes this the rare IT audit a CFO should personally request.
How sprawl happens
It happens innocently. Licenses get bought for a department that reorganizes. Seats get assigned to roles that never had an AI-shaped workflow. The pilot group's enthusiasm gets extrapolated to the whole company. And because AI spend often lives in a different budget line than the productivity suite, the renewal sails through procurement without the scrutiny a new purchase would get. None of this is anyone's failure — it's what happens when adoption is measured by seats assigned instead of value produced.
What the telemetry actually shows
Microsoft 365 exposes usage data that most organizations never open: who is actively using Copilot features, in which apps, how often, and who hasn't touched them since the seat was assigned. When that report finally gets pulled, the shape is remarkably consistent — a minority of power users generating real value, a middle band of occasional use, and a long tail of seats that have been idle for months. Industry adoption studies have found the same pattern across deployments of every size. Your tenant's exact numbers are an afternoon of work to surface; the willingness to look is the scarce ingredient.
The audit, step by step
A proper license audit does four things: pulls per-user activity telemetry across a meaningful window (90 days, not 7); maps assignments against roles to find structural mismatches — seats where the job simply doesn't contain Copilot-shaped work; identifies the power users and what they're doing, because their workflows are your internal playbook; and prices the reallocation — which seats to reclaim, reassign, or convert to a lighter license tier. The audit routinely pays for itself within the first quarter, and it's a fixed-fee project inside our Cloud & Microsoft 365 practice.
Right-sizing without a revolt
The mistake is treating this purely as cost-cutting. Yank licenses abruptly and you teach the organization that using new tools gets things taken away. The better sequence: celebrate and study the power users, offer training to the middle band before reclaiming anything, and reassign idle seats toward people whose workflows match what the power users are doing. Spend goes down and value per seat goes up — which is a very different board slide than "we cut licenses."
The bigger picture
License telemetry is also a readiness signal: it tells you where AI-shaped work already lives in your company, which is exactly the evidence an AI readiness assessment builds on — and keeping license governance continuous, rather than a one-time cleanup, is a standing line item in every Managed AI Operations bundle. If your renewal is coming up, book a call before you sign it. If the deeper question is who should run all of this, the managed-vs-hire math is the companion read.