GitHub Copilot Admin Guide

Last updated: 2026-06-18

A practical, task-first guide for a GitHub organization or enterprise admin managing GitHub Copilot.

This guide is designed for actual use, not passive reading. Start with the short path below, then jump into the card set that matches the job or problem in front of you.

Start Here

If you are setting up Copilot admin controls for the first time, use this order:

1. Assign Copilot access.
2. Create a universal user-level budget.
3. Add power-user overrides if needed.
4. Create an organization or enterprise spending limit.
5. Review AI usage and confirm the numbers you are looking at.
6. Export usage data when you need a deeper review.
7. Restrict premium models if burn is too high or governance requires it.

Pick Your Path

If you need to...	Go to this set
Roll out Copilot with basic guardrails	Set 1: Start and Guardrails
Monitor burn and understand what the UI means	Set 2: Monitor and Interpret
Pull data for finance, analysis, or investigation	Set 3: Export and API
Explain a mismatch or blocking issue	Set 4: Troubleshoot
Control expensive or restricted models	Set 5: Model Controls
Run Copilot as an ongoing admin practice	Set 6: Operating Routine

What Matters Most

Keep these four numbers separate at all times:

Metric	Meaning
Live AI usage	Full month-to-date usage
Budget consumed amount	Usage counted against a specific budget record
Included pool remaining	Shared credits left before overage
Metered usage	Spend after the included pool is exhausted

Most confusion comes from mixing them.

Admin Access You Need

You typically need one of these:

• Organization owner
• Enterprise owner
• Billing manager

For API access, you also need the right billing permissions or token scopes.

Set 1: Start and Guardrails

Card: Assign Copilot Access

Type: Action
Use when: You are enabling Copilot for a team, organization, or enterprise.

Do this

1. Go to your organization or enterprise settings.
2. Open Copilot or Billing & Licensing.
3. Assign Copilot access to the intended users.
4. Confirm the number of active licenses.

Watch for

• Included AI credits are pooled across the billing entity.
• For June to August 2026, promotional AI credit amounts may apply.
• Code completions remain included and do not consume AI credits.

Next

Go to Create a Universal User Budget.

Card: Create a Universal User Budget

Type: Action
Use when: You want a default guardrail for all users.

UI path

Billing & Licensing -> Budgets and alerts -> New budget

Choose

• Budget type: Bundled AI credits budget
• Budget scope: Users
• Leave the user blank for a universal budget

Suggested starting values

User type	Suggested budget
Standard user	$30 to $50
Heavy user or agent workflows	$75 to $200
Restricted user	$0 to $10

Watch for

• User-level budgets are intended to be the first budget check.
• User-level budgets are supposed to hard-stop usage when exhausted.
• Usage before budget creation may not count toward the budget for the current cycle.

Next

Go to Add Power-User Overrides or Create a Spending Limit.

Card: Add Power-User Overrides

Type: Action
Use when: A few users need materially higher limits than the default budget.

Do this

1. Create another Bundled AI credits budget.
2. Keep the scope as Users.
3. Select a specific user instead of leaving the user blank.
4. Set a higher limit based on actual usage.

Good use cases

• Approved agent-heavy workflows
• Admins or developers running evaluations
• Teams allowed to use premium models more heavily

Watch for

• Do not create exceptions before you have at least a little usage data unless there is a business need.
• Too many overrides make budgets harder to reason about.

Next

Go to Create a Spending Limit.

Card: Create a Spending Limit

Type: Action
Use when: You need an organization, cost center, or enterprise cap after the included pool is exhausted.

Do this

1. Create a new Bundled AI credits budget.
2. Choose Organization, Cost center, or Enterprise scope.
3. Set a dollar limit.
4. Enable Stop usage when budget limit is reached.
5. Add alert recipients.

Watch for

• If Stop usage when budget limit is reached is not enabled, the budget can act more like an alert than a guardrail.
• Spending limits and user-level budgets do different jobs. Do not treat one as a substitute for the other.

Next

Go to Read the AI Usage Dashboard.

Set 2: Monitor and Interpret

Card: Read the AI Usage Dashboard

Type: Action
Use when: You need a live operational view of current month usage.

Use this screen to see

• Total AI credits consumed
• Usage by user
• Usage by model
• Included credits remaining
• Additional usage or overage

What to watch for

Signal	Meaning
High AI usage by one user	Likely agent-heavy or premium-model-heavy usage
Net amount still zero	Shared included pool is still covering usage
Metered charges appearing	Shared pool is exhausted and overage has started

Next

Go to Read the Budgets Page or Export Current Usage.

Card: Read the Budgets Page

Type: Insight
Use when: You need to know what a budget record is actually tracking.

Use this screen to see

• Budget amount
• Budget consumed amount
• Users near limits
• Whether users should be blocked at budget exhaustion

What matters

• The budget page reflects usage counted against a specific budget record.
• It may not equal the user's full month-to-date AI usage.

Watch for

• If a budget was created mid-cycle, earlier usage may not appear in consumed_amount.

Next

Go to Why AI Usage and Budget Usage Do Not Match.

Card: Understand the Four Key Numbers

Type: Insight
Use when: Someone is mixing usage, budget, included pool, and overage into one number.

Keep these separate

Metric	What it tells you
grossQuantity	AI credits used
grossAmount	Dollar-equivalent month-to-date AI usage
consumed_amount	Amount counted against a budget record
netAmount	Actual billed overage after included pool and discounts

Admin rule

When a user appears over budget, check these in order:

1. Live grossAmount from the AI usage API.
2. consumed_amount from the budget API.
3. Budget creation timing.
4. Whether the budget is user-scoped or org-scoped.

Next

Go to Investigate a User Over Budget.

Set 3: Export and API

Card: Export Current Usage

Type: Action
Use when: You want quick analysis in a spreadsheet or need a shareable report.

Do this

1. Open Billing & Licensing.
2. Go to AI usage.
3. Click Get usage report or export the current view.
4. Download the CSV when it is ready.

Best for

• Quick monthly analysis
• Sharing with finance or operations
• Ranking users or models in a spreadsheet

Next

Go to Pull Live AI Usage by API if you need more precise or filtered checks.

Card: Pull Live AI Usage by API

Type: Action
Use when: You need current month live usage, per-user checks, or model-level analysis.

Example

gh api -H "X-GitHub-Api-Version: 2026-03-10" \
  "/organizations/ORG/settings/billing/ai_credit/usage?year=2026&month=6&user=USERNAME"

Read these fields first

• grossQuantity
• grossAmount
• netAmount

Best for

• Current month live usage
• Per-user monitoring
• Model-level analysis

Next

Go to Pull Budget Records by API when the budget page and usage page appear inconsistent.

Card: Pull Budget Records by API

Type: Action
Use when: You need to inspect configured budgets or explain what is counted against them.

Example

gh api -H "X-GitHub-Api-Version: 2026-03-10" \
  "/organizations/ORG/settings/billing/budgets?per_page=100"

Key fields

Field	Meaning
budget_amount	Configured budget cap
consumed_amount	Amount counted against that budget
prevent_further_usage	Hard-stop behavior flag
budget_scope	user, organization, enterprise, or cost center

Next

Go to Why AI Usage and Budget Usage Do Not Match.

Card: Pull Copilot Usage Summary

Type: Reference
Use when: You need a higher-level billing summary for Copilot usage.

Example

gh api -H "X-GitHub-Api-Version: 2026-03-10" \
  "/organizations/ORG/settings/billing/usage/summary?year=2026&month=6&product=Copilot"

Use this for

• Higher-level billing review
• Confirming whether overage has started
• Cross-checking monthly summaries

Set 4: Troubleshoot

Card: Why AI Usage and Budget Usage Do Not Match

Type: Insight
Use when: The budget page and AI usage page both look right, but they do not agree.

What is usually happening

• The AI usage API usually shows full month-to-date usage.
• The budget API can show only what is counted against a specific budget record.
• If a budget was created mid-cycle, older usage may appear in AI usage but not in the budget's consumed_amount.

Check now

1. Pull live grossAmount from the AI usage API.
2. Pull consumed_amount from the budget API.
3. Check when the budget became active.
4. Confirm the budget scope.

Next

Go to Investigate a User Over Budget.

Card: Investigate a User Over Budget

Type: Action
Use when: A user appears to be far beyond their budget or the user says the numbers make no sense.

Check in this order

1. Live grossAmount from the AI usage API.
2. Budget consumed_amount from the budget API.
3. Budget creation timing.
4. Whether the budget is user-level or another scope.
5. Whether the user is using a feature that does not consume AI credits.

Common explanation

• The budget was created after part of the user's monthly usage had already happened.

Next

Go to Check Why a User Is Not Blocked.

Card: Check Why a User Is Not Blocked

Type: Action
Use when: AI usage looks high but the user is still able to continue.

Check

1. Was the budget created mid-cycle?
2. What is the budget consumed_amount?
3. Is the budget user-scoped or another scope?
4. Is the user using a model or feature that is not consuming AI credits?
5. Does the budget record have prevent_further_usage enabled where relevant?

Watch for

• High AI usage alone is not sufficient proof that the blocking rule should already have fired.

Card: Check Why Usage Continues After a Spending Limit

Type: Action
Use when: Organization, enterprise, or cost-center usage continues after a limit appears reached.

Check

• Whether Stop usage when budget limit is reached is enabled on the relevant budget.

Common explanation

• The budget is acting as an alert because hard stop is not enabled.

Set 5: Model Controls

Card: Decide Whether to Restrict Models

Type: Decision
Use when: Burn is rising, governance is tightening, or you need a safer default rollout.

Restrict models when

• Premium frontier models are driving unexpected burn.
• Security or compliance requires tighter control.
• You want a standard baseline for most users.

Do not overcomplicate this

• Start with a small approved model set.
• Expand only after you understand real usage.

Next

Go to Apply a Model Policy.

Card: Apply a Model Policy

Type: Action
Use when: You are ready to enable only the model set you want users to access.

Do this

1. Open organization or enterprise Copilot settings.
2. Open model controls or model policy settings.
3. Review the list of enabled models.
4. Disable models you do not want available.
5. Keep only the approved set for your environment.

Good first pass

Scenario	Suggested model policy
General rollout	Enable standard lower-cost models first
Controlled agent usage	Enable premium models only for approved power users
Tight budget period	Disable highest-cost models temporarily
Regulated environment	Allow only approved compliant models

Set 6: Operating Routine

Card: Daily Checks

Type: Action
Use when: Burn is high or the rollout is still early.

Do this

• Check the AI usage dashboard if burn is high.
• Review top users and remaining included credits.

Card: Weekly Review

Type: Action
Use when: You are actively tuning rollout guardrails.

Do this

• Export usage data.
• Review model mix.
• Adjust individual user budgets.
• Check for users unexpectedly blocked or unexpectedly not blocked.

Card: Monthly Reset

Type: Action
Use when: You want to re-baseline normal usage and refine policy.

Do this

• Re-baseline normal usage.
• Tune universal and individual budgets.
• Review overage settings and guardrails.
• Reassess model availability.

Recommended Starter Configuration

For a new Copilot admin rollout:

• Universal user-level budget: $30 to $50
• Individual power-user overrides: $75 to $200
• Organization or enterprise spending limit: set explicitly
• Stop usage when budget limit is reached: enabled for spending limits
• Premium models: limited to approved users or teams
• Weekly CSV or API review: mandatory for the first month

Final Admin Checklist

• Copilot licenses assigned
• Universal user-level budget created
• Power-user overrides created
• Organization or enterprise spending limit created
• Hard stop enabled where applicable
• Alerting enabled
• Model availability reviewed
• CSV export process tested
• API extraction tested
• Budget consumed versus live usage understood by admins