The Microsoft Copilot usage report helps an administrator assess adoption. It is not a live prompt monitor, a transcript export, or proof that employees completed their work successfully. To avoid misleading conclusions, record the report, time window, and metric before comparing results.
Official reporting guidance checked October 14, 2026. This article concerns organizational Copilot reporting, not a personal account’s chat history.
Open the correct report with an appropriate role
Use an approved reporting role such as Reports Reader. Microsoft’s usage-report overview lists the permitted roles and cloud availability. Some roles show summaries without user details; do not request Global Administrator just to inspect adoption.
- Sign in to the Microsoft 365 admin center for the intended tenant.
- Open Reports > Usage. If Reports is hidden in the navigation, select Show all.
- Choose Microsoft Copilot, then Copilot, and open its Usage tab.
- Choose the available period and record it with the export. The current Copilot usage documentation lists 7, 28, 90, or 180 days; do not assume its monthly window is identical to a different report’s 30-day window.
Follow the Copilot-specific report guide if your tenant’s navigation differs. Copilot usage, Copilot Chat usage, agent reporting, and Viva Insights dashboards are separate views.
Read the metrics without mixing their meanings
Enabled users is the licensing population; active users is the population that intentionally used an in-scope capability. Prompts submitted counts activity, not people. Last activity is a UTC date, not a transcript or a daily total.
For example, 20 people producing 100 prompts is not 100 active users. Adding app-level user counts can count the same person more than once. A rise in enabled seats can lower an active-user rate even when the number of active people stays unchanged.
Use these questions when presenting a result:
- Is the population everyone licensed during the period, the current license holders, or a selected cohort?
- Are the two exports using identical periods, apps, and metric definitions?
- Does a blank value mean unavailable data, no recorded activity, or a field not included in this view?
- Has the pilot gained or lost users since the previous measurement?
Do not interpret a high prompt count as quality or a low count as failure. A worker who completes a task with one accurate prompt may obtain more value than someone repeating a failed prompt many times. Pair adoption metrics with voluntary task feedback and verified outcomes.
Export a reproducible comparison
The report provides chart exports and a user-table Export control. Use Choose columns where available, and keep the downloaded CSV’s column names rather than renaming different metrics to “interactions.” User-specific information is hidden by default; retain the organization’s privacy settings unless an authorized owner approves a change.
- Save the original CSV in an approved restricted location.
- Record the tenant, report name, selected period, retrieval time, and the report’s latest data date.
- Work on a copy. Keep the original values, including blanks, distinguishable from calculated zeros.
- Compare like-for-like periods and user populations. If you filter a cohort after export, label that analysis as your own filtered calculation.
- Share only the detail needed for the decision; an adoption presentation usually does not need everyone’s names.
This export contains reported usage data, not every underlying prompt and answer. Use the relevant, separately authorized compliance workflow for content investigation. Do not reconstruct “official usage” from audit events and assume the resulting counts must match.
Handle missing or delayed data safely
The Copilot-specific guide currently says activity typically appears within 48 hours after the end of the UTC day. Broader Microsoft reporting guidance allows longer delays. Recent numbers can also be revised. This is not a real-time dashboard, so do not treat today’s blank cell as proof that nobody used Copilot.
The report is absent
Check the signed-in tenant, reporting role, exact report, subscription, and cloud availability. Lack of access to a report does not establish that no Copilot license exists. Ask the administrator to verify those items before changing products.
A licensed user’s activity is missing
Confirm that the user intentionally used an in-scope feature and compare its timestamp with the report’s data date. Opening a pane alone is not a useful activation test. Keep a harmless, documented test event and allow the reporting delay before escalating.
Exports disagree
Compare time zones, periods, populations, refresh dates, and metric definitions first. If a discrepancy remains, keep the original exports and report it through the organization’s support channel. Do not enable optional diagnostics or weaken network controls just because an adoption cell is empty.