What organizations struggle to answer about Copilot
The Copilot questions that reach an executive team are rarely about activity counts. They are about whether a large, renewable commitment is earning its place:
- How many of the seats we bought are genuinely in use?
- Which departments adopted Copilot, and which never started?
- Is usage deepening over time or plateauing after the launch?
- What is Copilot costing us per active user?
- How does Copilot compare with the other AI tools we pay for?
- What do we take into the renewal conversation?
Some of that is answerable inside Microsoft. The rest requires joining Microsoft data to contract terms and to platforms that live entirely outside the tenant.
What Microsoft provides — and where the boundary sits
Microsoft publishes genuinely useful Copilot reporting. The Microsoft 365 admin center and the Graph Copilot reporting endpoints show enabled and active users, activity trends over standard periods, and per-app last-activity signals across the Microsoft 365 suite. For understanding what happens inside Microsoft 365, it is the authoritative source, and Midgentic treats it that way — the connector reads exactly those endpoints.
The boundary is scope rather than quality. Microsoft reports on Microsoft. It does not know what your organization pays per seat under your specific agreement, what your engineering teams spend on a coding assistant, what a business unit is running on a model API, or what value your organization has agreed to attribute to time saved. Those are portfolio questions, and they are answered one layer up.
What Midgentic actually ingests from Microsoft
Only what the Graph Copilot reporting endpoints expose to your tenant. Nothing here is inferred or reconstructed:
- Copilot user counts and trends
- Enabled and active Copilot user counts over the reporting periods Microsoft exposes, kept as a historical series rather than a snapshot.
- Per-user Copilot activity
- The usage user detail report, including last-activity dates that indicate whether a licensed person is actually using Copilot.
- Per-app activity signals
- Copilot activity across Microsoft Teams, Word, Excel, PowerPoint, Outlook, OneNote, and Loop, so adoption can be read by surface rather than as one number.
- Seat data
- Copilot seat counts as reported by Microsoft, which is what utilization is measured against.
- User identity resolution
- Microsoft users are matched to people in your workspace so activity can be attributed to departments and teams.
- License terms (entered by you)
- Contracted seats, price, currency, and effective dates. Microsoft's reporting endpoints do not expose billing, so this is entered rather than synced.
Prompt content, message bodies, and document contents are never collected. Midgentic reports on activity signals, not on what anyone wrote.
The questions this data answers
- Adoption
- How many licensed people use Copilot, how that is trending, and which departments are ahead or behind.
- Utilization
- Assigned seats against genuinely active users — the gap that matters most before a renewal.
- Usage patterns
- Which Microsoft 365 surfaces adoption actually landed in, which is usually where enablement should focus next.
- License context
- Contracted Copilot seats and terms held alongside the activity behind them, in the currency you are billed in.
- Spend and value context
- Copilot cost per active user, and Copilot's share of total AI investment, with estimated value kept separate from measured outcomes.
- Executive reporting
- A Copilot narrative that sits inside a portfolio report rather than as a standalone vendor export.
One platform inside a portfolio view
Midgentic is not a replacement for Microsoft's reporting and does not attempt to reproduce it. It is the layer above: it takes Copilot adoption as one input, adds the other AI platforms you run, adds the license and value data that no vendor holds, and produces the cross-vendor view that Copilot alone cannot support.
In practice that means Copilot adoption feeds adoption analytics, Copilot license terms feed spend intelligence, and both feed the AI ROI picture your finance team eventually asks for.
Common questions
What Microsoft 365 Copilot data does Midgentic collect?
Midgentic uses the Microsoft Graph Copilot reporting endpoints available to your tenant: Copilot user count summaries and trends, and Copilot usage user detail, which includes last-activity dates per Microsoft 365 app such as Teams, Word, Excel, PowerPoint, Outlook, OneNote, and Loop. It also resolves users so activity can be attributed to departments.
Does Midgentic show Microsoft 365 Copilot cost automatically?
No. Microsoft's Copilot reporting endpoints expose activity, not billing. Copilot license cost comes from license terms you enter — seats, price, currency, and effective dates — which Midgentic then reads against the activity data to show utilization and cost per active user.
How is this different from the Microsoft 365 admin center Copilot report?
Microsoft's reporting is the authoritative source for what happens inside Microsoft 365, and Midgentic uses it. The difference is scope: Midgentic places Copilot alongside every other AI platform and manually tracked tool in your portfolio, joins it to license terms and value assumptions, and produces one executive view across vendors.
Does Midgentic read Copilot prompts or document contents?
No. Only the aggregate and per-user activity signals exposed by the Microsoft Graph reporting endpoints are collected. No prompts, responses, files, or message contents are read.
Is the Microsoft 365 Copilot connector generally available?
It is implemented and in beta. It syncs real tenant data and is enabled per workspace during onboarding while it is validated with early customers.