Microsoft Copilot Cowork is now generally available — and it introduces a new way to pay: usage-based billing. Here's what your team needs to know.
Unlike a typical assistant that drafts and recommends, Cowork completes whole tasks end-to-end. Ask it to prep for a client meeting, and it can analyze your emails, CRM, and notes, build a briefing doc and a presentation, and email your team a summary — all from one prompt. It runs securely in the cloud, even when your laptop is off.
Standard Microsoft 365 Copilot is a flat monthly fee. Cowork uses a consumption-based pricing model because simple tasks and complex analyses require different levels of computing resources. You'll need a Microsoft 365 Copilot license first, then you're billed for Cowork based on the tasks you run.
More details on each below.
Our recommendation: Start every Cowork (or broader Copilot Credit) deployment on pay-as-you-go for one to two billing cycles. Use the Cost Management dashboard and Microsoft's Customer Cowork Estimator to establish a real consumption baseline. Once you've seen at least 60 days of steady-state usage:
Important: Capacity Packs ($200 / 25,000 credits/month) are not currently a billing option for Cowork. They apply only to Copilot Studio, Copilot Chat, SharePoint, and Dynamics 365 agents. If your tenant uses those agents alongside Cowork, capacity packs can offset that usage — but Cowork itself will always draw from P3 (if present) and then PAYG.
Coretek can help you analyze your tenant's Cowork telemetry and recommend the right blend.
For a more comprehensive overview, please review the information below.
A Copilot Credit is Microsoft's universal unit of measure for AI work — the way a utility meters electricity by the kilowatt-hour. Instead of paying a flat fee per task, you spend credits, and each task draws the number of credits it actually needs.
A few important things to know about how credits actually work:
1. One credit = $0.01 (one cent) at list price. 100 credits = $1. That's the pay-as-you-go rate. There are two ways to buy credits:
2. Credits are a shared, tenant-wide currency — not a per-seat allowance. There is no "X credits included per Cowork user." Your tenant has one pooled balance, governed centrally, and the same credits also pay for other Microsoft AI services that use this billing model — today that includes Copilot Cowork and the Work IQ API, with more agents and services being added over time (Copilot Studio custom agents, Dynamics 365 first-party agents, etc.).
3. Every task's credit cost is the sum of four things happening behind the scenes:
|
Cost driver |
What's actually being measured |
|
Models |
Which AI model ran the task and how much it had to reason. A frontier model (e.g., Claude Opus 4.8) costs more per step than a lighter model like the upcoming Cowork 1 |
|
Context |
How much of your data Cowork had to read and ground against — emails, files, chats, CRM records, meeting notes |
|
Tools |
Each action it took on your behalf — drafting a doc, building a deck, sending an email, calling a connector or plugin |
|
Runtime |
How long the orchestration ran end-to-end, including any retries or self-checks |
Source: Microsoft's GA pricing model.
4. Microsoft groups tasks into three rough tiers to make budgeting easier:
|
Tier |
Typical credits |
What it looks like |
Example |
|
Light |
~100–300 (Microsoft default: 125) |
Narrow context, lightweight model, 0–1 tool calls, 0–1 outputs |
"Summarize my inbox from the last 24 hours" |
|
Medium |
~400–700 (default: 500) |
Richer context, capable model, several tool calls, 2+ outputs |
"Draft a weekly status report from my Teams chats and project files" |
|
Heavy |
700+ (default: 2,500) |
Broad context aggregation, deep reasoning, many tool calls, many outputs |
"Analyze emails, CRM, notes, and support tickets for the Adatum account, then produce a briefing doc, a slide deck, and email a summary to the team" |
Defaults are based on Microsoft Frontier customer usage as of 5/27/2026 and assume Anthropic Opus 4.8.
5. Heavy users can cost 3–4x light users. In Microsoft's own estimator, a manager profile lands around $120/user/month in credit consumption, while a technical worker profile can exceed $400/user/month — on top of the underlying Microsoft 365 Copilot license. Blended averages run around $200–$240/user/month in credits before any P3 discount.
6. You control where credits come from and who can spend them. All of this is governed from the Microsoft 365 admin center → Copilot → Cost Management dashboard, where administrators set the billing method — pay-as-you-go or a P3 Pre-Purchase Plan — allocate credits, define spending policies, and apply hard caps and alerts at the tenant, group, or user level.
In one sentence: A Copilot Credit is the penny-priced "meter reading" for agentic AI work — pooled across your tenant, consumed by four measurable factors, and fully governed from a single admin dashboard so a variable bill never becomes an unpredictable one.
|
Tier |
Credits |
List price |
Discount |
Effective Cost |
|
1 |
300,000 |
$3,000 |
5% |
$2,850 |
|
2 |
1,500,000 |
$15,000 |
6% |
$14,100 |
|
3 |
3,000,000 |
$30,000 |
7% |
$27,900 |
|
4 |
15,000,000 |
$150,000 |
8% |
$138,000 |
|
5 |
30,000,000 |
$300,000 |
10% |
$270,000 |
|
6 |
75,000,000 |
$750,000 |
12% |
$660,000 |
|
7 |
150,000,000 |
$1,500,000 |
14% |
$1,290,000 |
|
8 |
225,000,000 |
$2,250,000 |
17% |
$1,867,500 |
|
9 |
300,000,000 |
$3,000,000 |
20% |
$2,400,000 |
Go to Copilot and then the Cost Management node to get started with usage-based billing

Because credits are a single, tenant-wide currency, Cowork is only one of several Microsoft AI experiences that draw from the same pool. As Microsoft expands its agentic platform, those same credits increasingly pay for work happening well beyond Cowork:
Work IQ APIs — When developers tap Microsoft’s workplace-intelligence layer directly through its APIs, usage is billed in credits: a variable charge for query-style work (grounding, retrieval, and reasoning), plus a flat 0.1 credit per call for the Tools API that takes actions on a user’s behalf.
Microsoft Copilot Studio agents — Custom agents you build and deploy consume credits as they retrieve information, respond to prompts, and run workflows or AI tools on a user’s behalf — including voice agents, which are billed by total call length.
Dynamics 365 agents — Both prebuilt and custom agents across Dynamics 365 draw from the same tenant credit pool, allocated to environments to manage usage across workloads.
Power Platform (Power Apps & Power Automate) — Credits apply only when AI or agentic features are invoked — generative AI, summarization, document processing, or an embedded Copilot Studio agent.
What doesn’t draw credits: the everyday Microsoft 365 Copilot experiences your per-user license already covers — Copilot Chat, the Copilot features in Word, Excel, PowerPoint, Outlook, and Teams, and Microsoft’s built-in agents like Researcher, Analyst, and Facilitator. Those stay on the flat monthly subscription. In short, credits meter the
autonomous, agentic work that runs on top of Copilot — and because every one of these services shares the same pooled balance, you can govern and budget all of it from the same admin dashboard.
Usage-based billing ties cost directly to value. The key is governance: turn it on deliberately, set limits early, and track ROI. Coretek can help you plan, configure, and budget for Cowork with confidence — reach out to your Customer Success team to get started.
Additional Links:
Copilot Cowork is now generally available | Microsoft 365 Blog
Managing AI experiences enabled by usage-based billing | Microsoft Learn