Microsoft Finally Named the Meter
If you’ve tried to explain the cost model for Microsoft Copilot agents to a CFO or a procurement team, you know how awkward that conversation gets. The per-user license makes sense for employees using Copilot to draft emails and summarize meetings. But what about an automated agent that runs a workflow overnight, calls an API, queries a database, and sends a notification — all without a human anywhere in the loop? What user license covers that? How do you budget for it? How do you prevent runaway spend?
For the past year, the honest answer has been: it’s complicated, and the billing model is still evolving. Build 2026 changed that. Microsoft introduced Copilot Credits — the named consumption meter for agent workloads that fall outside the standard per-user license scope. The concept isn’t entirely new, but giving it a name, a clear trigger condition, and a defined home in the admin center is a meaningful step toward a model that organizations can actually plan around.
Here’s what you need to understand before your agents go live.
The Core Distinction: Included vs. Metered
The licensing logic for Copilot agents now splits cleanly into two categories. Understanding where that line sits is the most important thing an M365 admin or decision-maker can take away from Build 2026.
Included scenarios are covered by the M365 Copilot add-on license. These are in-Copilot experiences — agents and workflows that run through the standard Copilot surfaces like Teams, Outlook, Word, and the Copilot chat interface. Users with a qualifying M365 Copilot add-on license can call Work IQ APIs in these contexts at no additional cost.
Metered scenarios consume Copilot Credits. These kick in when you go outside the standard Copilot surfaces. Specifically: your own custom agents calling Microsoft IQ, third-party agents integrating with Microsoft IQ, and complex Copilot Studio flows with heavy external tool use.
The line in plain English: If a licensed user is interacting with Copilot through a Microsoft-built surface and the agent is using Work IQ as part of that experience, no credits are consumed. The moment an agent works autonomously — outside a user session, or built by you or a third party — the meter starts.
The Scenarios That Will Catch You Off Guard
The included vs. metered distinction sounds clean in documentation. In practice, there are several common deployment patterns that will generate Credit consumption faster than most organizations expect.
Overnight automation is the highest-risk scenario. An agent that monitors a queue, processes incoming requests, and routes work — running continuously without user interaction — is entirely in metered territory. Unlike a user sitting at a desk for eight hours, an automation agent can execute hundreds of actions per hour. Without consumption controls, a single well-intentioned automation can generate significant Credit spend in days.
Copilot Studio flows with multiple tool calls are a subtler version of the same problem. A flow that looks simple — user asks a question, agent responds — can include several background tool calls to Microsoft IQ, external APIs, and data sources. Each tool call has a cost. Complex flows compound quickly.
Third-party agent integrations are the scenario most likely to catch IT off guard during a procurement cycle. If a vendor’s product integrates with Microsoft IQ to surface workplace context, and that product is deployed organization-wide, the Credit consumption from that integration may not be visible in your standard Copilot billing view.
Watch for this: Always-on autonomous agents — like Microsoft Scout, covered in the companion article on Autopilots — represent the maximum-consumption scenario. An agent that monitors your entire organization’s inbox and Teams activity around the clock will generate Credit consumption at a rate that is very different from a user opening Copilot to draft one email.
What’s Available in the Admin Center
Copilot Credits appear in the Microsoft 365 admin center under the billing and subscriptions area. As of GA, organizations can view Credit balance, monitor consumption by agent or workflow type, and set alerts when consumption crosses defined thresholds.
What’s not yet fully available — and worth confirming with your Microsoft rep — is granular per-department or per-agent credit caps. The monitoring capability is present. Hard limits that automatically throttle a specific agent before it burns through its Credit allocation are still evolving. Until those controls are mature, the most reliable governance approach is design-time discipline: scope your agents narrowly, test consumption in staging before production, and build throttling logic into the agents themselves.
Five Questions to Ask at Your Next Microsoft Renewal
Copilot Credits will appear in procurement conversations sooner than most organizations are ready for them. These are the questions worth raising before you sign anything.
- Are Credits included in our current EA or MCA, or do they require a separate SKU? The answer varies by agreement type and timing. Don’t assume Credits are bundled with your existing Copilot seat licenses.
- What is the overage policy if we exceed our Credits pool? Understand whether Microsoft will throttle agents, bill overages automatically, or pause workflows. All three have operational implications.
- Can Credits be scoped or capped at the department, tenant, or agent level? This is the governance question. The answer today is partially yes, and the roadmap is toward more granular controls — but confirm current capability vs. planned capability.
- How will Credits interact with our Copilot for M365 seat count at renewal? As agent workloads grow, the balance between per-user licenses and Credits will shift. Make sure your renewal structure can accommodate that shift without forcing a renegotiation mid-term.
- What reporting is available for Credit consumption by vendor or integration? If you’re deploying third-party agents that call Microsoft IQ, you need to be able to attribute Credit spend to those integrations, not just see it as an undifferentiated total.
How to Prepare Before Your Agents Go Live
The organizations that handle Copilot Credits well are the ones that treat consumption management as a first-class part of agent design — not an afterthought that gets addressed after the first surprise billing cycle.
- Classify every agent you’re building or deploying as included or metered before it goes to production. Make this a required step in your agent development checklist.
- Run consumption monitoring in staging. Microsoft provides tooling to estimate Credit usage before an agent hits production traffic. Use it.
- Build a Credit budget into your M365 cost model for FY27. Even a rough estimate is better than arriving at your next renewal with no data to negotiate against.
- Assign a Credit owner. Someone needs to own the consumption dashboard the same way someone owns the Azure cost management view. If nobody owns it, nobody notices until it’s a problem.
The Bottom Line
Copilot Credits are a reasonable answer to a real billing problem. Microsoft had to find a way to meter agent workloads that don’t map cleanly to per-user licenses, and a named consumption unit is a better model than the vague “it depends” that preceded it.
The risk is not the model itself — it’s deploying agents without understanding it. The organizations that will handle this well are the ones that read the meter before they turn on the tap. Get your admin center monitoring configured, update your procurement conversations, and make sure every agent in your environment has a clear answer to the question: included or metered?