The purchase came before the framing
The licence arrived with the suite, sometimes inside a wider negotiation. The use case was never defined up front, so nothing allows anyone to judge whether the tool delivers.
Platform · Microsoft 365 Copilot
Microsoft 365 Copilot is already deployed. How many seats produce?
Microsoft 365 Copilot is the most common entry point for enterprise AI: it arrives with a suite that is already installed, it scales by licence, and it is often decided before any usage framing. Six months later the question is always the same. Who uses it, for which tasks, and what does a seat return against what it costs.
Book a 30-minute discovery callIn short
On a Microsoft 365 Copilot estate, the admin center counts active seats, not what each seat produces or what happens in the other tools. An audit of real usage combines those reports with interviews to tell you which seats to renew, which to reassign and which uses to equip.
Which Copilot this page covers
This page covers Microsoft 365 Copilot, the assistant inside the productivity suite. GitHub Copilot, the developers' coding assistant, which companies buy as its Business and Enterprise plans, is also billed per seat but only handles code: see what usage changes in GitHub Copilot pricing.
Copilot Chat or a Microsoft 365 Copilot seat: what the seat pays for
In short
An eligible Microsoft 365 licence already gives Copilot Chat, at no extra cost. A Microsoft 365 Copilot seat, paid on top, mainly adds two things: reading the person's work data, and priority access to Copilot in the suite's apps. At renewal, the question becomes: which licence holders actually use what the seat adds?
| What changes | Copilot Chat, included in the licence | Microsoft 365 Copilot seat |
|---|---|---|
| The cost | Included with an eligible Microsoft 365 licence | An add-on paid per seat, on top of the suite licence, except in Microsoft 365 E7, which includes it |
| The data it reads | The web, and what the user gives it: an attached file, the open email or meeting | In addition, the person's work data, read through Microsoft Graph within their permissions |
| Word, Excel, PowerPoint and OneNote | No access with Copilot Chat alone. Microsoft provides standard access, subject to service capacity, for some users without the add-on | Priority access, faster and steadier at peak times |
| Agents | Those grounded in instructions and public websites; the others, pay-as-you-go only | In addition, agents that read work data, part of whose usage is included in the seat, under conditions |
NoteBased on Microsoft's documentation, read on 6 November 2026. Microsoft is renaming its offers: depending on the page, the add-on licence is now called Microsoft Copilot or Microsoft 365 Copilot (Premium), and the app Microsoft Copilot.
Markers
- 40+AI diagnostics delivered in a year
- 41,200users measured
- 20 to 40%of paid AI seats unused across the estates analysed (an engagement finding, not a study)
NoteCompany markers, across all our engagements: none is attributed to a client, none is a result promised in advance.
Why Microsoft 365 Copilot ends up on the table
In short
Copilot almost never raises a technical problem. It raises a steering problem, because the purchase decision precedes any knowledge of usage. Five situations recur.
Rolled out by population, not by usage
Seats were opened by department or by grade, rarely by task. Two people in the same team can have unrelated usages, and only one of them justifies the seat.
The renewal deadline
The contract is coming up and the question lands: how many seats to renew. Without a usage measurement, the answer is guesswork in one direction or the other.
A bill split across departments
The spend is spread across several budgets, sometimes across several entities. Nobody sees the full cost, so nobody challenges it.
Microsoft 365 Copilot is never alone
It almost always coexists with tools brought in by business teams. Comparing one use case from one tool to another is exactly what the vendor dashboard cannot produce.
What Copilot does inside an organisation
In short
On the Microsoft 365 Copilot estates we open, five usages recur: drafting and reworking documents, meetings and minutes, search across the document estate, agents built by teams, and data analysis in spreadsheets. For each one: what it is, what measurement reveals, and the decision it unlocks.
| Use | What it is | What measurement reveals | The decision it unlocks |
|---|---|---|---|
| Drafting and reworking documents | First drafts, rewriting, summarising long documents, formatting inside the office tools already in use. | Real frequency per person and per team, and above all depth: a usage that stops at the first draft does not return the same as one that runs through to the delivered version. That gap separates a seat that produces from a seat that is merely open. | Target training on the teams that stop at the first draft, rather than widening the estate. |
| Meetings and minutes | Transcription, summary, decisions and actions captured from meetings held in the conferencing tool. | Often the most regular usage and the easiest to attribute to a task. Measurement gives the share of meetings genuinely processed and what becomes of the minutes produced: read, circulated, or never reopened. | Industrialise the format where it gets reused, drop it where the minutes never leave the inbox. |
| Search across the document estate | Querying the organisation's documents, messages and shared spaces, within each person's permission scope. | Unserved demand: the questions asked that the tool cannot answer, usually because the source is not indexed or not current. That is a diagnosis of your document estate as much as of the tool. | Connect and clean the sources before widening. Success depends less on the model than on access to the right data. |
| Agents and instructions built by teams | Assistants configured by business teams for a specific task, with their own instructions and sources. | How many exist, which are published and which stay private, and above all the duplicates: the same need handled in parallel by teams unaware of each other. A shared asset reaches markedly more people than a personal version. | Publish across the organisation what is being rebuilt everywhere, and share agents on recurring needs. |
| Data analysis in spreadsheets | Exploring, shaping and interpreting datasets in the spreadsheet tool, for populations who are not analysts. | A usage heavily concentrated on a few people, with a per-seat return well above the estate average. These are the profiles that hit usage caps while other seats produce no trace at all. | Reallocate tiers between populations rather than buying. Often an arbitration at zero net cost. |
The five criteria that decide
In short
We do not publish a ranking between platforms. We publish the grid a leadership team uses to decide for itself, and what to look at in order to fill it in for Microsoft 365 Copilot specifically. The five criteria hold across all the platforms we measure; the middle column changes every time.
| Criterion | What to look at | What it tells you |
|---|---|---|
| Integration into the work environment | Does the work already happen inside Word, Excel, Outlook and Teams, or does the usage require opening a separate interface | This is Copilot's structural advantage, and the first driver of adoption, well ahead of model quality |
| Quality on the target task | The rework rate on your own documents, and the share of meeting notes that genuinely get reopened | A market average says nothing about your case |
| Governance and administration | Whether permissions genuinely hold when the assistant queries the document estate, and how the agents built by teams are administered | What conditions any extension beyond a pilot, especially if search surfaces poorly partitioned documents |
| Cost per seat against real usage | Share of active seats, depth of usage, and agent consumption, which is billed separately from the seat | The only criterion that turns spend into return |
| Sovereignty and hosting | Where processing happens, contractual commitments, retention policies applicable to interactions | Often an entry constraint, rarely a differentiator on its own |
What gets put in place once the measurement is done
In short
Once the measurement is done, three tooling moves follow, from the cheapest to the most committing: publish the assets, put AI where the work already happens, share agents. Measuring changes nothing on its own, and you do not fix a usage by asking teams to try harder.
Publish the assets
Agents, instructions and templates are created locally and stay private, so every team rebuilds the same ones. Publishing them across the organisation costs nothing and converges practice.
Put AI where the work already happens
Integration into the suite's tools is Copilot's structural advantage, and it only materialises if internal sources are connected and current. An assistant that cannot find the right source sends the user to another tool.
Share agents
On the needs measurement has shown to recur from one team to the next: a shared agent replaces as many private assistants. The most committing lever, and the only one that can be decided on data.
What actually drives the cost of a Microsoft 365 Copilot estate
In short
The real cost of a Microsoft 365 Copilot estate comes down to six mechanics: an add-on on top of a suite licence, an annual commitment, the active rather than the open seat, depth of usage, agents billed partly beside the seat and the hidden cost of rework. We publish no amount, neither ours nor a vendor's: a price list goes stale and says nothing about your situation; the mechanics decide.
| Mechanic | What it covers | Effect on the real cost |
|---|---|---|
| It is an add-on, not a standalone licence | Except in Microsoft 365 E7, which includes it, a Copilot seat sits on top of an eligible suite licence that must already exist. The real cost of an equipped user is therefore the add-on plus the licence that makes it possible, and the two do not always sit in the same budget. | Cost per equipped user exceeds the add-on's listed price |
| The commitment is annual | The common modality is a one-year commitment per seat. A seat opened mid-year stays paid until term, whether it produces or not. That is what makes the renewal date so structuring. | Waste is observed for twelve months before it can be corrected |
| The active seat, not the open seat | The console itself separates enabled users from active users. Dividing the spend by the latter rather than the former changes the order of magnitude of the unit cost. | Changes the order of magnitude of the unit cost |
| Depth, which the active count does not carry | A user counts as active as soon as they trigger a feature. Between someone who stops at the first draft and someone who goes through to the delivered version, the return is not remotely comparable, and the usage report does not separate them. | Two seats at the same cost for unrelated returns |
| Agents are billed partly beside the seat | Only part of agent usage is included in the seat: that of a licence holder who uses an agent under their own identity, in Copilot Chat, Teams or SharePoint, within fair-usage limits. Computer-using agents are not included, nor are flows started by any trigger other than an agent calling them: like the usage of people without a seat, they are settled in Copilot Credits, pay-as-you-go, through capacity bought in advance or a one-year commitment. An estate that industrialises its agents therefore sees a second line appear, on a different rhythm from the first. | A second spend line, usually discovered after the fact |
| The hidden cost of rework | Time spent correcting output appears on no invoice. On some usages it exceeds the time returned, and it is the line the console does not carry. | Can exceed the time returned |
NoteNo quote is given before the analysis. The real cost of your estate is computed on your own logs and presented in a meeting. Targeted and measured at your site: a 30 to 50 % reduction on the cost of the targeted usages, never the whole bill. That is a mission objective, never a result promised in advance.
The shape a seat estate takes
What your console shows, and what it does not
In short
You already have a usage report, and it is useful. So it is worth saying exactly what it carries before saying what escapes it. Everything below is checkable in Microsoft's own documentation, and our article on the Copilot usage report details how it defines an active user.
What the data shows
- Enabled users and active users: How many seats are assigned, how many triggered a feature, and the rate that follows.
- The split by application: Which suite applications carry the usage: word processing, spreadsheets, mail, conferencing.
- The most used agents: A ranking of the agents called, and activity per agent and per user.
- Prompt volume: How many prompts were submitted and the average per user, as totals and as a daily series.
- Four reading windows: Views over 7, 28, 90 or 180 days, usually available within 24 to 72 hours.
What it does not show
- Depth of usage: A user is active from the first prompt. Nothing separates the person who stops at the first draft from the one who goes through to the delivered version.
- What becomes of the output: A set of meeting notes produced is counted. Whether it was read, circulated, or never reopened is not.
- Rework time: Time spent correcting what the model produces appears nowhere, although it decides the real return.
- What happens elsewhere: By construction the console covers the Microsoft perimeter. It cannot see the teams who switched to another tool, and therefore allows no comparison of one use case from one platform to another.
- Duplicates across teams: The agent ranking gives the most called, not the same need rebuilt in parallel by five teams unaware of each other.
- Attribution, when anonymisation is on: When reports conceal user names, which Microsoft does by default, usage can no longer be tied to a team. The setting protects and blinds the steering at the same time.
NoteNone of this is a flaw in the product: a vendor instruments its own perimeter, as expected. What we add is the view across every tool and the reading of what happens after the output, from the same exports you already administer.
How a Copilot estate gets measured
In short
Copilot is never alone in an organisation. The measurement that counts is the one that places it among the other AI tools on the estate, because that is the only way to compare one use case from one tool to another.
The logs and exports you already produce
We start from what the platforms you administer already emit: Copilot, but also internal models, other assistants and custom agents. Nothing installed on workstations, no re-instrumentation. Export handover or read-only connection, the exact modality is settled in the scoping meeting.
The view across every tool
A vendor dashboard covers its own vendor's perimeter, by construction. It does not see the people who switched elsewhere, nor the same need handled on two platforms. That consolidated view is precisely what allows arbitration: Idun Cockpit, our tool, builds it from your logs, and you can keep it after the engagement, on subscription, to follow it over time.
Interviews, for what logs do not say
Logs give the facts: who uses what, how often, at what cost, with what error rate. They do not give intent. Interviews with the IT department and the business units are a mandatory step. The interviews and the data, never one in place of the other.
A read-out that holds up in front of finance
The document reads without reprocessing. Finance finds the full cost and the return per seat, engineering the detail per tool, HR the training need that follows.
The measurement loop
Three deliverables
The map
What teams actually do with Copilot and with the rest of the estate: detected use cases, real adoption per division, agents and instructions created, published or kept private.
The account
What each use case costs and what it returns: paid seats against active seats, return per seat, caps reached, consolidated full cost.
The course
What to industrialise first, which seats to renew or reallocate at renewal, who to train on what.
Usage, never people
We analyse usage, never people. Results are aggregated per team and per tool, and the exact processing status is validated with your data protection officer. Read-only, nothing written back into your systems, no workstation scanning. The report also carries its appendix of unknowns: the perimeter actually covered, what we could not see, and the figures we refuse to put forward.
What we have already measured
In short
Our measured, publishable case is a consulting and training firm equipped with Claude. Idun Group does tooled consulting and resells no platform: we have to be able to conclude that the tool is not worth what it costs at your site.
Consulting and training firm
At a consulting and training firm, the analysis of real Claude usage established 85% real activation across around forty seats, false starts included. With the seats that left no trace at all, 31% of seats, nearly one in three, were to be reassigned or activated. At the firm, the three heaviest users' consumption is worth, at API prices, 12 to 25 times the price of a premium seat.
The limitThe limit, written at the top and never in an appendix: the platform export had a rolling thirty-day retention while the rollout had been running for five months. One month out of five was measured.
NoteThe estate measured was not a Copilot estate, and that is precisely the point: the mechanics measurement surfaces do not depend on the platform, they depend on what you can read.
What leadership teams ask us
What does Microsoft 365 Copilot cost for an enterprise?
We publish no tariff, neither ours nor a vendor's: a price list goes stale and says nothing about your situation. What decides is cost against real usage. A paid seat that goes unused costs full price for zero return, and that is what gets measured.
Which licence does Microsoft 365 Copilot require?
A Microsoft 365 Copilot seat is an add-on: it sits on top of an eligible Microsoft 365 or Office 365 licence, such as E3, E5, Business Standard or Business Premium. Only the Microsoft 365 E7 suite includes it. The Microsoft Copilot Business offer is reserved for organisations of up to 300 users.
Is Copilot Chat enough, or do we need Microsoft 365 Copilot seats?
Copilot Chat, included with eligible Microsoft 365 licences, is enough for someone who works from the web and from the content they give it. The paid seat is justified for someone who needs Copilot to read their work data and to assist them with priority access in Word, Excel, PowerPoint and OneNote. The split between the two is decided on each team's real usage, not on the org chart.
How do we know how many of our Microsoft 365 Copilot seats genuinely serve?
By comparing open seats with the seats that produce a trace, then looking at depth of usage: an active seat is not necessarily a seat that produces.
Is the Microsoft dashboard not enough?
It covers its own vendor's perimeter, by construction. It does not see people who switched to other tools, nor the same use case handled on two platforms. That comparison is what allows arbitration.
Should we pick Microsoft 365 Copilot over another platform?
We publish no ranking. We publish the five criteria a leadership team uses to decide for itself, and we fill them in with your own data rather than a market average.
Are you a Microsoft reseller?
No. Idun Group is an independent consultancy and resells no platform. Our measurement tool, Idun Cockpit, can be kept after the engagement, on subscription: it measures the platforms and replaces none of them.
Do we need to install anything on our workstations?
Nothing. We work from existing exports and logs, read-only, with no re-instrumentation. Processing runs in France, at Scaleway in Paris, or inside your perimeter if your policy requires it.
Will our teams feel individually monitored?
No. It is a method rule: we measure usage aggregated per team and per tool, never a named employee's activity. We do not look for who uses the tool badly, but for where the tool is badly set up.
What is the difference between Microsoft 365 Copilot and GitHub Copilot?
They are two distinct products, both billed per seat. Microsoft 365 Copilot is the assistant inside the productivity suite, used across every function. GitHub Copilot is the coding assistant, used by engineering teams. This page covers only the first.
How many of your Microsoft 365 Copilot seats actually produce?
Thirty minutes to scope the perimeter, the platforms to read and the renewal deadline you have in mind. You leave with a straight verdict on whether the exercise makes sense at your scale.
Book a 30-minute discovery callRead-only. We analyse usage, never people.
Supported by
