---
title: "Paid Per Seat, Used Per Task"
source: https://bitvaria.com/en/bezahlt-pro-lizenz-genutzt-pro-aufgabe
---

30\. Juli 2026 · [AI](https://bitvaria.com/category/ai/)  · 7 min read

# Paid Per Seat, Used Per Task

300 paid Copilot seats are not 300 users. What Microsoft's own reports count as "active," what the widely cited 20-30 percent figure actually measures, and why it likely understates the gap rather than overstates it.

![300 paid Copilot seats are not 300 users. What Microsoft's own reports count as "active," what the widely cited 20-30 percent figure actually measures, and why it likely understates the gap rather than overstates it.](https://bitvaria.com/_astro/bezahlt-pro-lizenz-genutzt-pro-aufgabe.CQk8m9at.jpg)

> **Paid per seat, used per task.**
> 
> A seat costs the same whether it gets used or not. By most reported figures, only 20 to 30 percent of paid Microsoft 365 Copilot seats are weekly active, and even “active” is generously defined: a single intentional action a week is enough. Your own admin report measures with the same definition. Whether a seat gets work done or merely gets touched once a week is not something any standard report shows. It can still be measured, from your own usage data.

Vendors are happy to report sold seats, in millions and with growth rates. Used seats, nobody reports. And on the invoice, both cost the same: per head, per month, up front.

## Two ledgers, one gap

Microsoft, like most AI assistant vendors, sells per head, per month. The price applies whether the person opens the seat once a day or once a quarter. That is not an accident, it is the business model: margin comes from the gap between the flat fee and actual consumption. Lightly used seats are the most profitable ones for the vendor.

The alternative would be usage-based billing, the way the underlying model APIs work. No waste from non-use, but variable cost from month to month. Almost every company still buys seats, not consumption, because flat fees are easier to budget. That same simplicity is exactly what hides the gap.

## What gets reported

On the sales side, the number is clear: in January 2026, Microsoft reported 15 million paid M365 Copilot seats worldwide, up 160 percent year over year. And as early as late 2024, by Microsoft’s own account, nearly 70 percent of the Fortune 500 were using Copilot.

On the usage side, it gets blurrier. Press reports put the share of weekly-active seats among paid ones at 20 to 30 percent; no audited primary study backs that range. According to a Recon Analytics survey, Copilot’s share among paid AI subscribers in the US fell from 18.8 percent in July 2025 to 11.5 percent by January 2026. Both usage figures are press and analyst reporting, not first-party measurement, and should be flagged as such in any conversation that cites them; all sources are listed at the end of this post.

Microsoft’s own behavior indirectly confirms the problem is real: in December 2025, the company cut prices in the SMB segment, and “Frontier Company,” a multibillion-dollar, several-thousand-person program, is now running to move customers from pilot into real usage. A vendor does not fund a program that size against a problem it considers small.

## What “active” actually means

The M365 admin center report under Reports → Usage → Copilot shows Enabled Users against Active Users, per app, with windows from 7 to 180 days. The definition of “active” is the weakest one available: at least one intentional Copilot action in the chosen window. A single such action counts. Whether it turned into a completed task is not something the report shows.

Usage is more useful thought of as a four-rung ladder: enabled, then active by Microsoft’s definition (at least once a month), then habitual (weekly), then value-generating. The widely cited 20 to 30 percent sits on the third rung from the bottom. None of the publicly cited figures measure how much of that reaches the fourth rung. There is no reason to assume it is larger than the base it is built on.

## Active is not done

The active-user rate implies a simple state: a seat gets used, or it does not. Reality is more granular. A seat can be “active” and still create almost no value: one summary requested per week, the result skimmed, the task then done by hand as always. The report counts that as usage. Whether the action turned into a completed task is counted nowhere.

That has an uncomfortable consequence for any measure that targets the active-user rate alone. More clicks raise the rate without moving the real target, completed work. A number that looks better on paper can still hide the same gap it did before.

## What this means for your company

The reported 20 to 30 percent is an industry average in which a single action a week counts as usage, not a statement about any one company. Your own admin center report is the first honest step, because it replaces an industry estimate with your own number. It still rests on the same weak definition of usage, though: it shows who clicked. It does not show what share of those clicks turned into completed work, which of that is routine and would run just as well on a cheaper model, or what the same volume would cost on a usage-based model.

Without your own number, every conversation about AI license cost stays a conversation about someone else’s survey. With it, you can at least answer the first, most basic question: who opened the seat at all.

One question stays open here, and in many German companies it is the harder one: how to measure usage at all without sidelining works council co-determination. More on that in an upcoming post in this series.

## Frequently asked questions

**How many paid AI seats are actually used?** Press reports put the share of weekly-active seats among paid Microsoft 365 Copilot licenses at 20 to 30 percent. No audited measurement backs that range, and “active” here means only: at least one intentional action a week. A specific company’s number can sit above or below it and can only be determined from its own usage data.

**What counts as an “active” seat in Microsoft 365 Copilot?** In the admin center, a user counts as active once they have performed at least one intentional Copilot action within the chosen window, typically a rolling 7, 28, 90, or 180 days. A single intentional action within the window is enough for “active” status. The report says nothing about how often, for what, or with what outcome a seat gets used.

**Why isn’t the admin center report enough on its own?** The report shows who clicked. It does not show which tasks got done, what share of those are routine, what the same volume would cost on a usage-based model, or where the data actually flows. It measures the bottom rungs of a multi-step usage ladder, not real value.

## Sources

-   **15 million paid Copilot seats (January 2026):** Microsoft figure from the FY26 Q2 earnings call, documented by [Office 365 for IT Pros](https://office365itpros.com/2026/01/30/microsoft-fy26-q2-results/) and, in German, by [SharePoint360](https://sharepoint360.de/microsoft-verkuendet-15-millionen-abonnierte-m365-copilot-lizenzen/).
-   **Nearly 70 percent of the Fortune 500 (late 2024):** [Microsoft’s own Ignite announcement](https://blogs.microsoft.com/blog/2024/11/19/ignite-2024-why-nearly-70-of-the-fortune-500-now-use-microsoft-365-copilot/).
-   **20 to 30 percent weekly active:** widely cited press estimate with no audited primary study behind it, e.g. [ValueAdd VC](https://valueaddvc.com/blog/microsoft-copilot-enterprise-adoption-what-the-data-shows-about-real-usage-vs-hype).
-   **18.8 to 11.5 percent share among paid AI subscribers (US):** [Recon Analytics, “AI Choice 2026”](https://www.reconanalytics.com/ai-choice-2026-why-licenses-dont-equal-adoption/), survey of over 150,000 respondents.
-   **SMB price cut (December 2025):** [Computerworld](https://www.computerworld.com/article/4093224/).
-   **“Frontier Company” ($2.5 billion, around 6,000 staff):** [CNBC](https://www.cnbc.com/2026/07/02/microsoft-commits-2point5-billion-6000-employees-ai-implementation-unit.html) and [TechCrunch](https://techcrunch.com/2026/07/02/microsoft-launches-its-own-ai-deployment-company-with-2-5-billion-commitment/).
-   **Admin center report and the definition of “active”:** [Microsoft Learn](https://learn.microsoft.com/en-us/microsoft-365/admin/activity-reports/microsoft-365-copilot-usage).

* * *

*Related: [Residency Is Not Jurisdiction](https://bitvaria.com/en/residenz-ist-nicht-jurisdiktion) and [Hosting or Renting Your Own AI Models](https://bitvaria.com/en/ki-modelle-hosten-oder-mieten).*

-   [ki](https://bitvaria.com/tag/ki/)
-   [mittelstand](https://bitvaria.com/tag/mittelstand/)
-   [copilot](https://bitvaria.com/tag/copilot/)
-   [lizenzkosten](https://bitvaria.com/tag/lizenzkosten/)
-   [nutzungsmessung](https://bitvaria.com/tag/nutzungsmessung/)

Share:

[Back to Blog](https://bitvaria.com/blog/)