How Much Does Microsoft Copilot Cowork Actually Cost?
Short answer: Microsoft Copilot Cowork — a different product from Anthropic's own Claude Cowork, worth keeping straight since both share the name — became generally available worldwide on June 16, 2026, at $30 per user per month for the required Microsoft 365 Copilot seat license. That license doesn't include any actual agent work: every task Cowork runs is billed separately in "Copilot Credits" at $0.01 each, with Microsoft's own published tiers putting a light task around $1–$3, a medium task at $4–$7, and a heavy multi-step task at $7 or more — no included allowance, no ceiling. A single heavy task can already cost more than a third of bots.team's entire flat $19/month Pro tier. One clarification worth making up front: Microsoft's $30 seat covers the AI itself, while bots.team's app tiers price the app only — you bring your own Claude subscription (from $20/month) to actually power your bots, so the fair comparison is $30/month-and-rising against a flat $20–$39/month all-in.
Cowork's pitch is simple — describe the outcome you want and the agent plans, executes, and revises the work. The bill for that work is where it gets complicated.
What Copilot Cowork actually is
Cowork is Microsoft's agentic layer inside Microsoft 365 Copilot — it went generally available worldwide on June 16, 2026, after a roughly three-month preview. Instead of answering one question at a time, you describe an outcome ("pull last quarter's numbers into a deck") and Cowork plans the steps, works across Word, Excel, PowerPoint, Outlook, and Teams, and delivers a finished result. It's built for the same "describe it once, let it work" idea bots.team is built around — the difference that actually matters to your monthly bill is how each one charges for that work.
The two-layer pricing model, in plain numbers
Cowork's pricing has two separate parts, and it's easy to see only the first one in a headline:
- The seat license — $30/user/month. This is the required Microsoft 365 Copilot license. Without it, Cowork isn't available to that user at all. This fee buys access, not usage.
- Copilot Credits — $0.01 each, metered separately. Every time Cowork actually does something — drafts a document, builds a spreadsheet model, researches something, sends messages — it consumes credits based on the model used, how much context it has to retrieve, how many tools it calls, and how long it runs. There's no bundled monthly allotment; it's pay-as-you-go unless an organization pre-commits to a volume discount.
| Task size | Approx. credit cost | What that buys, roughly |
|---|---|---|
| Light task | $1 – $3 | A short draft, a quick lookup, a small formatting pass |
| Medium task | $4 – $7 | A multi-step task pulling from a few sources or tools |
| Heavy task | $7+ | A long-running, multi-tool job — research plus a deck plus a summary email |
Run even a handful of medium tasks in a week and the metered portion alone can pass the $30 seat fee. Microsoft's own admin guidance for Cowork includes spending limits, budget allocation per user or group, and usage alerts precisely because the bill isn't predictable by default — those are cost-control tools for a cost that needs controlling.
Cowork is billed like a utility meter, not a subscription — the seat fee gets you access, and a separate dial keeps running underneath it.
What that looks like against a flat monthly price
bots.team's app charges $0 for the free tier (unlimited bots, unlimited runs), $5/month flat for Basic, or $19/month flat for Pro — that number doesn't move based on how many bots you run or how much work they do. But to be fair to the comparison: bots.team doesn't provide the AI itself, so that flat app fee isn't your whole bill. You also need your own Claude subscription to actually run bots — most people will want Claude Pro at $20/month, which itself is a flat fee (with generous but real session and weekly limits, not a per-task charge). So the real comparison is two flat numbers added together — bots.team's tier plus Claude Pro — against Cowork's uncapped, granular per-task meter on top of its seat license. All-in, that's roughly $20–$39/month for bots.team depending on tier, versus Cowork's $30/month seat before a single task has run.
| Copilot Cowork | bots.team Pro + Claude Pro | |
|---|---|---|
| Base price | $30/user/month (seat license, AI included) | $19/month app + $20/month Claude Pro = $39/month, both flat |
| Cost per task/run | Separately metered, $0.01/credit ($1–$7+ per task), no cap | Included in both flat fees, within Claude Pro's session/weekly usage limits |
| Can the bill change month to month? | Yes — scales with task volume and complexity, no ceiling | No, as long as usage stays within Claude Pro's limits — heavy users may need Claude Max ($100–$200/mo) |
| Needs admin spending controls to avoid surprise bills? | Yes — budgets, credit limits, usage alerts are part of setup | Not applicable — no per-task billing on either fee |
| Built for | Enterprise M365 tenants with IT admins managing licenses | Individuals, creators, and small teams managing their own bots |
This isn't a Cowork-is-bad point — it's a who-it's-built-for point. Metered, credit-based billing makes sense for an enterprise IT department managing hundreds of seats and wanting granular cost attribution per department. It's a much harder thing for a solo creator or small business owner to reason about without a finance team watching the meter for them.
Why this matters if you're not running an IT department
If you're a solo creator, a first-time founder, or someone managing social accounts for a small business, the honest question isn't "which agent is smarter" — it's "can I predict what this costs me in October." A metered system means the answer depends on how much you use it that month, and estimating that in advance requires understanding credit consumption per task type, something Microsoft's own pricing guides devote entire pages to explaining. A flat price means the answer is just the number on the plan.
To be fair about it: Cowork's usage-based model exists because the tasks it runs vary enormously in cost to actually execute — a document draft and a multi-tool research job are not the same amount of compute, and Microsoft is passing that difference through rather than averaging it into one number. bots.team's flat pricing works because its bots run on your own schedule at a pace you set, not on-demand enterprise-scale bursts — a different usage pattern, not a trick.
With metered billing, checking "how much did this cost me this month" means adding up individual task charges — something a flat monthly price never asks you to do.
What to actually check before picking a metered AI tool
- Is there an included allowance, or is every task billed separately? Cowork has none — 100% of task execution is metered on top of the seat fee. Some tools bundle a monthly credit allotment before metering kicks in; that's a materially different deal.
- Do you need someone watching a usage dashboard? If a tool's own documentation recommends setting spending limits and usage alerts, that's a signal the cost is variable enough to need active management — worth knowing before you commit, not after the first invoice.
- Does the pricing scale with your actual usage pattern? A tool built for occasional, scheduled tasks (check something once a day, once a week) has a very different cost shape than one built for constant, on-demand enterprise workloads — even if both are labeled "AI agents."
FAQ
Is Copilot Cowork included in a regular Microsoft 365 Copilot subscription? The $30/user/month Microsoft 365 Copilot license is required to use Cowork, but it doesn't include Cowork's task execution — that's metered separately via Copilot Credits, priced at $0.01 each, with no included allowance at general availability.
How much does a single Cowork task typically cost? Based on Microsoft's published credit tiers, a light task runs roughly $1–$3, a medium task $4–$7, and a heavy, multi-tool task $7 or more — all on top of the $30/month seat fee, per pay-as-you-go pricing.
Is "Copilot Cowork" the same thing as "Claude Cowork"? No — they're two different products that happen to share a name. Microsoft Copilot Cowork is the metered agent layer covered here, built into Microsoft 365. Claude Cowork is Anthropic's own product, built around Claude's Scheduled Tasks; it's priced differently and doesn't use a credit meter. Worth double-checking which one a source is actually talking about before comparing numbers.
Can I get a flat, predictable price for AI agent work instead? Tools built around flat monthly pricing exist for individuals and small teams rather than enterprise IT tenants — bots.team, for example, charges $0–$19/month flat for the app with no per-task metering. That app fee doesn't include the AI itself, though: you'll also want a Claude subscription (Claude Pro is $20/month, also flat) to run bots, so the realistic total is around $20–$39/month depending on tier — still two predictable numbers rather than a per-task credit meter.
Is metered pricing a bad sign? Not inherently — it fits enterprise IT teams who want granular cost attribution across departments and are already staffed to manage it. It's worth flagging mainly if you're evaluating agent tools without that kind of support and assumed the sticker price was the whole story.
Knowing the number in advance is its own kind of feature — one less thing to check before you commit to a tool.
If you're a solo creator or small-team owner deciding between the two, the deciding question isn't which agent is more capable — it's whether you want a bill you can predict before the month starts. bots.team's app tier is flat no matter how many bots you run or how often they run, and every run lands in an organized dashboard — reports, a discovery feed, metric history — that you can scan at a glance instead of reconstructing from a credit statement. Add your own Claude subscription and the whole setup stays two flat numbers, known in advance, for work that runs unattended on your own schedule.