Microsoft Turns On Copilot Usage-Based Billing by Default (Oct 19 / Nov 2, 2026): The 12-Step Spending-Policy Playbook for IT and Finance

What changed, in one paragraph

On September 21, 2026, Microsoft published Message Center post MC1476200, “Simplifying AI access: introducing usage-based billing.” The change is narrow but consequential: for new eligible Microsoft 365 Copilot Business purchases, Microsoft will automatically create the Azure resources that usage-based billing requires and apply a spending policy with default limits, so that credit-metered experiences such as Copilot Cowork and the Work IQ API work on day one. Direct Microsoft purchases flip first (October 19); the Cloud Solution Provider channel follows (November 2). Microsoft's own admin guidance is to review the eligible experiences, then “confirm or adjust usage controls and spending policies” in the Microsoft 365 admin center on or after the rollout date. Disabling usage-based billing is possible, but it removes access to the experiences that depend on it.

The partner announcement adds one detail buyers should hear plainly: the default-on configuration ships with the bundle purchases as well as standalone seats, and it includes access to Copilot Cowork, the Work IQ APIs, and GitHub Copilot Harness. In other words, the seat you thought was a fixed line item at Microsoft's $30 list price now has a variable tail unless you cap it.

Two meters, not one: what is actually billed by usage

Most of the confusion EPC Group sees in Copilot cost reviews comes from treating “pay-as-you-go” as a single bucket. Microsoft runs at least two distinct usage-billing systems for Copilot, and they are governed in different places.

1. Copilot Credits (Microsoft 365 admin center → Copilot → Cost management). Credits are the currency for Copilot Cowork, apps built with Cowork, and the Work IQ API, with Microsoft stating it will add more agents and services over time. Microsoft's published pay-as-you-go rate at GA was $0.01 per credit, with prepaid options at a discount. Cowork itself charges by model use, context retrieval, tool calls, and runtime, and Microsoft describes three task-complexity tiers (light, medium, heavy). Cowork is disabled by default and must be enabled and assigned by an admin, which is exactly why MC1476200 matters: the new default is that the billing plumbing exists the moment you buy, so an admin who enables Cowork for “a pilot group” without touching the policy is running on Microsoft's default limits, not yours.

2. Copilot Studio message meter (Azure subscription meters). Copilot Chat, SharePoint agents, and any custom agent or Copilot built in Copilot Studio bill at Microsoft's published $0.01 per message through the Copilot Studio meter, and the charge lands on an Azure subscription. These are managed in a different admin surface (Copilot → Billing & usage, plus Azure Cost Management), and they have their own budget mechanics. One line in Microsoft's budget documentation deserves to be read twice: for these pay-as-you-go services, reaching 100% of your budget does not stop the service or the billing — the only way to stop billing is to disconnect the service. Budgets alert; they do not enforce.

A third family sits alongside these and is billed through the Billing node rather than Copilot: Microsoft 365 Backup, SharePoint Storage pay-as-you-go, High Volume Email, and Microsoft 365 Archive. If your organization is turning on the Purview archive action for SharePoint files this fall (rolling out early October through early November 2026), that consumption shows up there, not under Copilot.

Why E7 does not make this go away

Microsoft 365 E7 — generally available since May 1, 2026 at Microsoft's list price of $99 per user per month with Teams — bundles E5, Microsoft 365 Copilot, Agent 365, and the full Entra Suite. It removes the Copilot seat line from your invoice. It does not remove the runtime line: Microsoft's own E7 page notes that agent work requires an Azure subscription and is “priced on a metered basis.” The same is true of Copilot Studio capacity (Microsoft sells it as $200 capacity packs of 25,000 credits or at $0.01 per credit pay-as-you-go) and of Cowork. Treat E7 as a seat-consolidation decision and usage-based billing as a separate FinOps decision. Organizations that conflate them discover the difference in the second month.

What a “default spending policy” actually does

When you activate the default spending policy, you are setting the tenant-level limit for all users. Every additional policy you create has its own independent limit and does not inherit the tenant-level one. Microsoft's setup flow exposes six decisions:

1. Billing method. Which Azure subscription pays for Copilot Credit usage. This becomes the default for every later policy.
2. Monthly policy limit. “Don't limit monthly spending” or “Limit monthly spending” to a credit ceiling.
3. Per-user monthly limit (optional, but Microsoft explicitly recommends it) to stop one person consuming the whole pool. When users hit their limit they lose access to the governed agents and services for the rest of the month until credits reset on the first.
4. Alerts. Email notifications when policy usage crosses a threshold (credit amount or percentage), repeating weekly until the period resets or you adjust the policy; optional user-level notifications when a per-user cap exists.
5. Auto-apply new services. On by default. Left on, every future Copilot service Microsoft adds to the credits model joins the policy automatically; turned off, you review and add each one manually.
6. Scope. The default targets the entire organization. You can narrow it to specific security groups before activating.

Three behaviors that surprise finance teams, all documented by Microsoft:

• Prepaid first. Policies always draw down prepaid credits (capacity packs, pre-purchase plans) before pay-as-you-go.
• Users in multiple policies are assigned by a fixed precedence: highest per-user limit, then largest overall policy limit, then most recently created. Settings are not combined.
• Group moves carry history. A user who moves between Entra ID groups mid-month keeps their consumption; they cannot “reset” by changing departments. A running task that exceeds the per-user cap can complete, and Microsoft states the excess is not billed and not counted.

Consumption reporting refreshes every two hours, can be viewed by policy, user, group, agent, and service, and — important for governance — Microsoft supports reader roles so finance and licensing staff can review spend without permission to change policies.

The 12-step spending-policy playbook

This is the sequence EPC Group uses with clients preparing for the October and November cut-overs. Steps 1–4 happen before you buy or enable anything; 5–9 configure the controls; 10–12 make the model durable.

1. Inventory the purchase path. Confirm whether your Copilot Business seats renew or expand through direct Microsoft billing (October 19 trigger) or a CSP (November 2 trigger). Ask your CSP in writing whether new orders will arrive with usage-based billing enabled and what default limits Microsoft applies. Existing subscriptions are unaffected; new SKUs and add-on purchases are not.
2. Name the owner. Usage-based billing sits at the seam between the Microsoft 365 admin and finance. Pick one accountable owner and one reviewer with a reader role.
3. Decide the Cowork posture before the plumbing exists. Cowork is off by default. Decide now whether the answer for the first 90 days is “off,” “pilot group only,” or “broad,” because the moment billing is default-on, an admin toggling Cowork is also authorizing spend.
4. Classify data readiness. Cowork inherits sensitivity labels end to end and is covered by audit logs, DSPM, eDiscovery, Insider Risk Management, and Data Lifecycle Management. That is only protective if the labels exist. If your tenant has not completed an oversharing and labeling pass, the spending policy is not your biggest risk. (This is the surface a Copilot readiness assessment is for.)
5. Replace the Microsoft default with your default. Activate the default policy with Limit monthly spending and a real ceiling derived from your pilot budget, not “don't limit.” Keep the scope to all users only if step 3 was “broad.”
6. Set the per-user cap. Microsoft recommends it; EPC Group considers it mandatory. Derive it from the task-tier economics: a heavy Cowork task can run into the hundreds of credits, so a per-user cap that allows a handful of heavy tasks a week is a reasonable first month for a pilot.
7. Create group policies for the pilot. Because specific users can only be targeted through security groups, create a Copilot-Cowork-Pilot security group, give it its own policy with a higher per-user cap than the tenant default, and remember the precedence rule (highest per-user limit wins). Do not put pilot users in two policies and expect the limits to add.
8. Wire alerts to two inboxes. One policy-level alert to the IT owner at 50% and 80%; one to finance at 80% and 100%. If a per-user cap exists, turn on user-level notifications so people are warned before they lose access on the 20th of the month.
9. Decide “Auto-apply new services” deliberately. For regulated organizations EPC Group recommends off for the first two quarters: every new metered service Microsoft adds should be reviewed for data handling before it inherits a budget. For everyone else, on is acceptable if step 8 is in place.
10. Separate the Copilot Studio budget. Agents built in Copilot Studio, SharePoint agents, and Copilot Chat are metered at Microsoft's published $0.01 per message on Azure, and their budgets alert but do not stop billing. Create the Azure budget, set alert thresholds at 50/80/100%, and define the disconnect procedure in your runbook, because disconnecting is the only hard stop.
11. Build the monthly review. The Consumption tab answers “where are we spending and are we on track” by policy, user, group, agent, and service. Export it monthly (the export is a point-in-time snapshot and can differ slightly from the live dashboard). Compare credits to outcomes: which agents earned their spend.
12. Put it in the chargeback model. Usage-based Copilot spend belongs in the same showback or chargeback framework as Azure. If departments do not see their agent runtime, they will not govern it.

A worked example (illustrative, using Microsoft's published rates)

A 1,200-seat organization buys 200 new Copilot Business seats through its CSP on November 10. Billing arrives default-on. IT enables Cowork for a 40-person pilot. If each pilot user runs three medium tasks a day for 20 working days and a medium task lands mid-range at roughly 500 credits, the month's Cowork runtime is 40 × 60 × 500 = 1,200,000 credits, or $12,000 at Microsoft's published $0.01 per credit — on top of the seats. A per-user cap of 15,000 credits a month holds the pilot to $6,000 at that same Microsoft rate and forces a conversation about which tasks are worth heavy runtime. The numbers are illustrative; the point is that the cap, not the seat price, determines the invoice.

Frequently asked questions