Copilot Credits spending-limit calculator: credits per user from Microsoft's published rates
EPC Group built this calculator so an IT or finance owner can set the organization's own Copilot spending limit before Microsoft's default applies: enter users and an expected monthly mix, and it returns credits per unlicensed user, the tenant total, the position against the 4,000-credit default and the 125% enforcement line — every rate Microsoft's, attributed below.
Last updated by Errin O'Connor, Founder & Chief AI Architect, EPC Group
Copilot Credits spending-limit calculator
Enter your users and an expected monthly mix per unlicensed user. The result is credits per user per month, the tenant total, where that sits against Microsoft's 4,000-credit default spending limit, and where the 125% enforcement line would sit for a prepaid balance. All arithmetic runs in your browser.
Reading the result. At this mix one unlicensed user's month is well inside the 4,000-credit per-user default Microsoft sets for new Copilot Business licenses (8% of it), a scale for the per-user limit you set rather than a limit Microsoft applies to agent users automatically. Setting the organization's own limit at or near the computed figure keeps the policy tied to a mix you chose rather than to the default.
Enforcement line. Enter a prepaid balance to see where Microsoft's 125%-of-prepaid enforcement would sit for this tenant.
Rates: Microsoft Learn, “Copilot Credits billing rates” (requirements-messages-management), read 2026-11-03; default limit: Microsoft Learn, Partner Center announcements, October 2026; enforcement threshold: the same rates page; monthly capacity: Microsoft Learn, Standard harness licensing. Links are below the tool. This is arithmetic on your inputs, not a quote.
The rates this calculator uses, and where each one comes from
| Agent feature | Copilot Credits | Microsoft 365 Copilot-licensed user |
|---|---|---|
| Classic answer | 1 | No charge |
| Generative answer | 2 | No charge |
| Agent action | 5 | No charge |
| Tenant graph grounding for messages | 10 | No charge |
| Agent flow actions (per 100 actions) | 13 | No charge |
| Content processing (per page) | 8 | No charge |
| Voice, per minute: classic · GenAI · premium GenAI | 10 · 35 · 75 | Included |
| Text and generative AI tools (premium): reasoning-model premium, per 1,000 billed tokens, on top of the feature rate | 10 | No charge |
Source for every row: learn.microsoft.com — Copilot Credits billing rates (requirements-messages-management), first read 2026-10-07 and re-read 2026-11-03 (page dated 2026-10-30), when the nine feature rates were unchanged. The same page states: “Enforcement is triggered when a tenant reaches 125% of their prepaid capacity” and “When enforcement applies, the environment's agents are blocked.” Default spending limit of 4,000 Copilot Credits per user per month for new Microsoft 365 Copilot Business licenses purchased through CSP from December 1, 2026: learn.microsoft.com — Partner Center announcements, October 2026, re-read 2026-11-03. Capacity enforced monthly, unused credits not carried over: learn.microsoft.com — Standard harness licensing. Billing methods and the “Auto-apply new services” default: learn.microsoft.com — usage-based billing overview. Capacity allocation and how enforcement is resolved: learn.microsoft.com — manage Copilot Credits capacity. If Microsoft changes a rate, this page is updated and the change is dated here.
How to read the result
A worked example. 500 users, 100 of them licensed for Microsoft 365 Copilot; each of the other 400 asks 20 classic answers, 60 generative answers, 15 agent actions, 10 tenant-graph-grounded answers and triggers 50 agent flow actions a month, with 25,000 prepaid credits a month (the tenant size in Microsoft's own enforcement example). Credits per unlicensed user: 322 a month (8% of the 4,000 default). Tenant total: 128,600 credits a month. At an even daily rate that total uses up the month's capacity on day 6, when new agent flow runs are blocked, and crosses the 125% line of 31,250 credits on day 8, when the environment's agents are blocked. That is the point: a default that is comfortable per user can still exhaust a small monthly capacity within days, and the limit the organization sets should be derived from its own mix, not inherited.
What to set the default to: a limit at or just above the per-user figure for your mix, assigned by group rather than to everyone, with the “Auto-apply new services” choice made deliberately for each policy — the twelve controls are in EPC Group's spending-policy playbook and the release that announced it, the Copilot usage-based billing default release. Agent-level governance — owners, scopes and the 125% behavior per environment — is in the nine Copilot Studio controls.
Four synthetic scenarios, worked
Invented mixes that show the arithmetic per unlicensed user per month; they are not client data or benchmarks. Each row, the worked example above and Microsoft's own billing examples (7,200, 4,800 and 20 credits, and 141 credits for one reasoning-model message) are checked by an automated arithmetic test before the page ships.
| Scenario (synthetic) | Mix per unlicensed user per month | Credits per user | Tenant per month | Against the 4,000 default | Monthly prepaid capacity |
|---|---|---|---|---|---|
| Public FAQ agent, 1,000 visitors | 30 classic and 10 generative answers | 50 | 50,000 | 1.3% | 200% of 25,000: capacity used up on day 15, agents blocked on day 19 |
| Employee agent grounded on the tenant graph, 150 employees (50 licensed) | 80 generative answers, each tenant-graph grounded | 960 | 96,000 | 24% | None entered |
| Support agent on a reasoning model, 200 users | 40 generative answers billing 14.1 thousand tokens each, 20 agent actions | 5,820 | 1,164,000 | 145.5% | None entered |
| Document intake agent, 50 users | 400 agent flow actions and 25 pages of content processing | 252 | 12,600 | 6.3% | 50.4% of 25,000: neither line is reached |
The assumptions behind the numbers, and when each was checked
- Rates. Microsoft's billing-rates table, first read 2026-10-07 and re-read 2026-11-03 (page dated 2026-10-30): the nine feature rates were unchanged and the reasoning-model premium was added. The calculator leaves out the basic and standard text and generative AI tool rates, and bring-your-own-model usage, which Microsoft bills separately.
- Licensed users cost nothing here. The table's “No charge” column covers employee-facing agents running under a Microsoft 365 Copilot-licensed user's signed-in identity. Its footnotes exclude agent flow runs started by any trigger other than “When an agent calls the flow” or by an autonomous agent, and Computer-Using Agents; enter that usage against unlicensed users.
- Prepaid capacity is monthly. Copilot Studio enforces purchased capacity monthly and unused credits do not carry over (Standard harness licensing, read 2026-11-03), so the calculator compares one month's tenant total with one month's capacity instead of drawing down a balance.
- Two lines in the month. When an environment has no credits left, new agent flow runs are blocked while the agent keeps answering; at 125% of prepaid capacity the environment's agents are blocked, including in a conversation already in progress. An environment that bills overage to pay-as-you-go is not blocked by enforcement (rates page, read 2026-11-03). The calculator places both lines at an even daily rate over a 30-day month, as if one environment drew on the whole capacity; environments with their own allocation are enforced separately.
- The 4,000 default. Microsoft's Partner Center announcement (re-read 2026-11-03) sets it for new Microsoft 365 Copilot Business licenses bought through CSP from December 1, 2026. Because Microsoft attaches it to Copilot Business licenses, it does not apply automatically to the unlicensed users of a Copilot Studio agent; the calculator shows it as a scale for the per-user limit you set.
- Reasoning models. Billed tokens are charged at 10 credits per 1,000 on top of the feature rate, and the count depends on the agent's instructions, knowledge, tools and conversation length; Microsoft's own illustrative follow-up message billed 14,100 tokens, or 141 credits. Measure your agent in a pilot before you enter a figure.
- No dollars. The calculator works in Copilot Credits only; Microsoft's pricing pages are the source for prices.
The spending worksheet, in short
- Count users and the licensed share; licensed users are not charged for these features.
- Estimate the monthly mix per unlicensed user from a pilot or from Copilot Studio's usage estimator.
- Multiply by Microsoft's rates (the table above); that is the per-user figure.
- Set the organization's own spending limit at or near it, by group, before usage-based billing turns on by default for your purchase channel (December 1, 2026 for new Copilot Business licenses through CSP).
- If capacity is prepaid, compare the monthly tenant total with the monthly capacity, place the 125% line, and decide per environment whether overage bills to pay-as-you-go.
- Record the policy, its owner and its limits in the AI governance register — surface eight of TAR-8.
Questions IT and finance teams ask
What is a Copilot Credit?
A Copilot Credit is Microsoft's unit for usage-billed Copilot and Copilot Studio features. Microsoft's billing-rates table (read 2026-11-03) prices a classic answer at 1 credit, a generative answer at 2, an agent action at 5, a tenant-graph-grounded answer at 10, agent flow actions at 13 per 100, content processing at 8 per page, and voice at 10, 35 or 75 per minute by tier; when a reasoning model orchestrates the agent, Microsoft adds 10 credits per 1,000 billed tokens on top of the feature rate. EPC Group's calculator uses those rates and nothing else.
Who is not charged Copilot Credits?
Microsoft's rates table marks these features "No charge" or "Included" for users with a Microsoft 365 Copilot license, for employee-facing agents that run under the licensed user's own signed-in identity. Its footnotes exclude agent flow runs started by any trigger other than "When an agent calls the flow" or by an autonomous agent, and Computer-Using Agents. EPC Group's calculator counts only the unlicensed users you enter, so enter that usage against them.
What is the 4,000-credit default?
Microsoft's Partner Center announcement for October 2026 (re-read 2026-11-03) states that new Microsoft 365 Copilot Business licenses purchased through CSP get usage-based billing enabled by default from December 1, 2026 (moved from November 2), with a default spending limit of 4,000 Copilot Credits per user per month that administrators can adjust. For direct purchases, Microsoft's message center post MC1476200 (September 2026) gave October 19, 2026; check the tenant's Message center for the current date. The default belongs to Copilot Business licenses, so it is not applied automatically to the unlicensed users of a Copilot Studio agent. EPC Group's playbook sets the organization's own limit before the default applies.
What happens at 125%?
Microsoft's rates page (re-read 2026-11-03) states that enforcement is triggered when a tenant reaches 125% of its prepaid capacity, that the environment's agents are then blocked, and that a blocked agent stops responding, including in a conversation already in progress; the designated administrator is notified by email. An environment that bills overage to a pay-as-you-go plan is not blocked by enforcement, and new agent flow runs are blocked separately, as soon as an environment has no credits left. Capacity is enforced monthly and unused credits do not carry over. EPC Group's calculator shows the day of the month each line is reached for the monthly prepaid capacity you enter.
Is this a price quote?
No. It is arithmetic on your own inputs using Microsoft's published credit rates. EPC Group publishes no dollar figure here; Microsoft's pricing pages are the only source for prices.
What does EPC Group do with the result?
EPC Group's Microsoft 365 Copilot Readiness Assessment scores the tenant on the eight TAR-8 surfaces, including cost governance, and writes the spending policy, its owner and its limits into the tenant's AI governance register before usage-based billing is switched on.
Where EPC Group fits
EPC Group has led 300+ Microsoft Copilot initiatives (rollouts and readiness assessments) across 11,000+ enterprise engagements since 1997; its Microsoft 365 Copilot Readiness Assessment scores the tenant on the eight TAR-8 surfaces — cost governance included — and writes the spending policy into the register before usage-based billing is switched on; the Microsoft 365 Copilot consulting practice carries the rollout from there. Organizations that want the limit set for their tenant can request a scoping conversation at epcgroup.net/contact.