The call
Turn on Stop usage when budget limit is reached for every spending limit.
Set a universal user-level budget above the per-seat value, and fund each business unit from its own seats through cost centers. Restrict frontier models to the organizations that need them, and choose who pays for code review before members without a license can request it.
Copilot keeps billing past your seat-funded credit pool by default, so one heavy user or agent session can spend the month's pool in a week and keep going. Since June 2026, every feature except completions and next edit suggestions draws on those AI credits. This checklist sets the budget controls in the order GitHub recommends, sizes them with GitHub's own method, and treats model and code review policy as part of the bill.
When you need this
- You buy Copilot Business or Enterprise seats and have not revisited budget settings since the move to AI credits.
- Agent use is growing and finance wants a ceiling on monthly Copilot spend.
- One team's agent sessions are using credits that other teams' seats paid for.
- Copilot code review is enabled, or about to be, for members without a Copilot license.
Evaluation criteria
Decide whether paid usage is allowed
When the pool runs out, should work stop or continue at US$0.01 per credit?
Look forChoose whether to enable AI credits paid usage in AI Controls. Organizations and enterprises have it on by default. Turning it off blocks usage once the pool runs out until it resets at 00:00 UTC on the first of the month. Unused credits do not roll over.
Set a universal user-level budget
What is the most any one licensed user should consume in a month, from the pool and paid usage together?
Look forSet a universal user-level budget above US$19 for Business or US$39 for Enterprise. That lets heavier users draw on unused credits from others. User-level budgets always stop usage at the limit. A US$0 budget blocks the user immediately.
Give named power users their own limit
Which engineers run long agent sessions or work in large codebases, and for how long do they need more?
Look forSet individual budget overrides for heavy users in the AI usage dashboard. Add expiration dates when the higher limit covers one sprint or incident.
Turn spending limits into hard stops
Does reaching your enterprise or cost center limit stop charges, or only send a notification?
Look forEnable Stop usage when budget limit is reached on every enterprise and cost center spending limit. Otherwise, GitHub sends alerts while charges continue.
Size the limits with GitHub's method
If every user spent their full user-level budget, how much would land as paid usage?
Look forSubtract the pool's value from maximum user-level consumption. In GILD's calculation with GitHub's method, 100 Business seats fund 190,000 credits worth US$1,900 a month. A US$40 user budget allows US$4,000 of total consumption. Spending limits must then permit US$2,100 of paid usage, or they block users before their individual budgets do.
Fund each business unit from its own seats
Should one team's agent use draw on credits another team's seats paid for?
Look forAssign users directly to cost centers. Set included usage controls so each unit draws no more credits than its licenses fund. Add a budget per cost center and an enterprise budget as a final limit. Organization budgets enforce unpredictably for users with licenses from several organizations.
Treat model policy as cost policy
Which models may each organization use, given the price spread between them?
Look forSet model policies per organization using the price table. Per million output tokens, GPT-6 Luna costs US$0.50. Claude Opus 5.5 costs US$20 and GPT-6 Astra US$50. Copilot does not switch to a cheaper model when a budget runs out. GitHub retires models on announced dates: six on Oct 19, 2026, and Claude Opus 4.8, Claude Opus 5 and GPT-5.6 Terra on Nov 16, 2026. Replacements turn on automatically unless an admin has turned off the global default or disabled the model. Organizations that did must enable Claude Opus 5.5, GPT-6.1 Sol or GPT-6 Luna in model policies before that date.
Decide who pays for code review
Who pays for each review: the pull request author, the organization, or no user budget at all?
Look forChoose the billing source for code review. Automatic reviews default to the pull request author. Reviews for unlicensed members bill the organization as paid usage outside user budgets. Reviews on GitHub-hosted runners also consume Actions minutes. Self-hosted runners consume none.
Review the bill monthly
Does the pool last the month without blocking anyone early?
Look forCheck the AI usage dashboard monthly for users blocked early and unexpected charges. Filter code-review minutes by copilot-pull-request-reviewer. Search action:copilot in the audit log for policy changes.
Questions to ask vendors
- What were our AI credit use and paid usage for each billing cycle since June 2026, by cost center and by model?
- Which users draw most from the pool, and through which features, such as chat, Copilot CLI, the cloud agent, or code review?
- How much of our code review spend bills the organization for members without a license?
- Which minimum IDE, extension, and Copilot CLI versions do our developers need for accurate usage and pricing display?
- Can volume pricing on AI credits be included in our GitHub or Microsoft agreement?
Red flags
- The AI credits paid usage policy left at its default with no enterprise spending limit.
- Spending limits without Stop usage when budget limit is reached, which only send alerts.
- A universal user-level budget set exactly at the per-license value, which stops pooling from helping heavier users.
- Organization budgets in an enterprise where users hold licenses from several organizations.
- Every frontier model enabled for every organization with nobody watching spend by model.
Products in this piece
Sources
- Official documentation
- 9
- Public source
- 3
- GitHub Copilot docs: Usage-based billing for organizations and enterprises
- GitHub Copilot docs: Getting started with budget controls
- GitHub Copilot docs: Budgets for usage-based billing
- GitHub Copilot docs: Optimizing your budget configuration
- GitHub Copilot docs: Models and pricing
- GitHub Copilot docs: Plans
- GitHub Copilot docs: Copilot code review
- GitHub Copilot docs: Policies for enterprises and organizations
- GitHub Copilot docs: Reviewing audit logs
- Hacker News: Kon5ole on a department's budget after the June 2026 change (Jul 2, 2026)
- Hacker News: K3UL on a company cost comparison after June 2026 (Jul 2, 2026)
- Hacker News: sandos on pooled credits across teams (Apr 28, 2026)