One shared credit pool. Pay-as-you-go beyond it. Four guards to stay in control — plus full visibility into who is using how much. This page explains how GitHub Copilot UBB works on Enterprise Cloud and exactly how TK Elevator should adjust cost centers, budgets, user segmentation and usage reporting.
These updates from the GitHub Changelog go beyond the original animated explainer — make sure your setup reflects them.
Every Copilot plan now bills based on AI Credits consumed. Included usage + optional additional spend with budgets.
Read changelog →Set a universal budget for all users or override for specific groups. For AI credits, ULBs control total usage, not just overage.
Docs →Doubled from 250 → 500. Map cost centers much closer to your real business-unit structure.
Read changelog →Programmatically create/update/delete budgets and pull usage by org, repo, cost center, product or SKU. Ideal for bulk ULB rollout.
API docs →New ai_credits_used field per user in the Copilot usage metrics API — connect consumption to value.
CCR consumes GitHub Actions minutes in addition to AI Credits. Configure a default org-level runner.
Docs →Set a default enterprise cap, then selectively grant additional usage via specific cost centers. Solves the "ringfencing" gap.
Read changelog →Workflows using GITHUB_TOKEN + copilot-requests: write bill the org directly — bypassing ULBs. Use cost-center budgets instead.
Manage cost-center assignments only via the dedicated Cost centers page in Billing settings (or API).
Read changelog →Every seat you buy adds AI Credits (AIC) to one shared pool for your enterprise. Everyone draws from the same jar. 1 AIC = $0.01 (100 AIC = $1).
Copilot Business: $19/seat → 1,900 AIC into the pool
Copilot Enterprise: $39/seat → 3,900 AIC into the pool
Example: 1,000 TK Elevator users on Enterprise = 3,900,000 AIC ($39,000) included every month.
Once your included credits run out, every additional AIC is billed at the end of the month — only up to the spending budget you set. No budget = no overage allowed.
Without it, the limit is just a notification — the meter keeps running and you can be billed for unlimited overage.
Two guards protect the jar (so one person can't drain it). Two guards protect your wallet (so you can't be billed beyond a cap).
Limits how many AIC one user can consume per month. Set a universal default or override for specific groups. GA since June 1, 2026.
Override the default ULB for specific users (e.g. platform engineers, AI champions) so they get a bigger slice without lifting the cap for everyone.
The maximum dollar amount your enterprise can be billed beyond included usage. New: you can now also create an enterprise budget that excludes cost-center usage, so specific business units can have their own headroom.
Group users, repos or orgs into cost centers (e.g. per TK Elevator business segment, region or product team) and cap each one. Enables chargeback, blocks one segment from burning the whole overage. Up to 500 cost centers per enterprise.
• No org-level ringfence of the included pool — heavy users in one org can drain shared credits other orgs were counting on. Workaround: cost-center budgets on the overage side.
• Agentic workflows billed to the org bypass ULBs. Control them via cost-center budgets + the agentic-workflows cost-management tools (gh.io/gh-aw-cost).
• A user / repo / org can be assigned to only ONE cost center at a time. Assigning to a new one auto-removes from the old one.
UBB only works if you can see consumption. GitHub offers four different surfaces for usage visibility — from a no-code dashboard to a full REST API. Here's where to look for what.
The fastest way to see active users, engaged users, suggestions accepted, IDE breakdown, language breakdown, Chat/PR-summary/agent activity for the last 28 days. Includes auto-resolved model names, Copilot CLI, cloud agent and plan mode.
The authoritative view of what you owe right now. Shows included vs additional AIC, dollar spend, budget consumption with progress bars, and breakdown by product (Copilot, Actions, Packages, Codespaces, GHAS).
github.com/organizations/YOUR-ORG/settings/billing.A row-per-event CSV for finance and chargeback. Every line tells you which user, which day, which product/SKU, how many AI Credits, and how many dollars. Now uses quantity (AIC) and gross_amount (USD) as the standard fields.
user to answer "who used the most?"cost_center for chargebacksku to see which premium models / features drove spendaic_quantity / aic_gross_amount preview columns are zero for AI-credit usage. Use quantity and gross_amount going forward.Three APIs cover the full visibility story. Combine them for "which user, in which team, in which cost center, burned how many AI Credits this month?"
Per-user adoption + AI credit consumption. NEW field: ai_credits_used (Jun 19, 2026).
GET /enterprises/{ent}/copilot/metrics
GET /enterprises/{ent}/copilot/usage/users-1-day
GET /enterprises/{ent}/copilot/usage/users-28-day
GET /enterprises/{ent}/copilot/usage/user-teams
Spend by year / month / day, filterable by org, repo, cost center, product, SKU.
GET /enterprises/{ent}/settings/billing/usage
GET /organizations/{org}/settings/billing/usage
Detailed line-item report (daily totals — hourly was removed in June 2026).
GET /enterprises/{ent}/settings/billing/usage
users-28-day from the metrics API → gives you ai_credits_used per user.user-teams → maps each user to their team(s).Metrics API → users-28-day, sort by ai_credits_used desc. Or AI usage CSV → pivot by user.
Metrics dashboard → Engaged users vs Active users. Or compare seat list vs users-28-day response.
Billing UI → Usage → filter by cost center. Or Usage Summary API with cost_center filter.
AI usage CSV → pivot by sku. Common SKUs: Copilot Chat, Agent, CCR, CLI, Spaces, premium model add-ons.
Billing UI → Budgets tab (live %). Or Usage Summary API with granularity=day → forecast.
AI usage CSV filtered to Copilot CLI SKU + the automation org. Plus gh.io/gh-aw-cost tools per workflow run.
Automated alert if any cost center crosses 75% of its budget (Usage Summary API + cron).
Quick check of the Billing UI Usage tab. Look for unusual spikes.
Pull the AI usage CSV, review top consumers, adjust ULBs, send chargeback report to BUs.
Re-baseline budgets with your account team based on 90-day trend. Right-size seats.
Most enterprises don't need flat ULBs. Segment your users by role, give each segment its own budget envelope and use cost centers to map to chargeback business units.
Combine ULBs (per-user cap) with cost centers (per-business-unit chargeback) for the strongest control.
Occasional Chat users, mostly inline completions. Business analysts, IT support, citizen devs.
ULB: ~$10 / moDaily Chat + occasional agent mode. Field-service app, web, ERP, modernization teams.
ULB: ~$30 / moHeavy agent use, CCR, CLI, Spaces. Digital Products (MAX), cloud platform, data engineering.
ULB: ~$80 / moOrg-billed agentic runs. ULBs don't apply — control via cost-center budget.
Cost-center capExact paths, what each setting does and the trade-offs — for cost centers, budgets, ULBs, enterprise limits and more.
A cost center is a group of users, repositories or organizations that you want to track and bill together — typically a business unit, brand, product or department. Cost centers are the foundation for chargeback and granular spend control.
Caps how much one user can consume. For AI Credits, this controls total usage — not just overage — so it directly protects the shared pool. Use a universal default + targeted overrides for power users.
Caps total additional (overage) spend for everything assigned to a cost center. The only way to ringfence overage spend per business unit and enable chargeback-style accountability.
The master ceiling on total additional spend for the whole enterprise. Now (preview) supports an option to exclude cost-center usage so you can set a "default-everyone-else" cap and let specific cost centers run with their own budgets.
The single most important setting in UBB. OFF = budget is just a notification, charges keep accruing. ON = hard stop, no surprise bills. Always ON unless you have a specific business reason.
Trigger emails at custom % thresholds. Always notify both the budget owner (e.g. team lead) and central billing managers, so nobody is surprised when the hard stop kicks in.
/enterprises/{enterprise}/settings/billing/budgetsManage the entire lifecycle of budgets via API — create, update, delete, change amount, change alerts. Required for any rollout at scale (hundreds of users / dozens of cost centers).
Since June 1, 2026, CCR consumes GitHub Actions minutes on top of AI Credits. You can now set a default Actions runner at the organization level instead of per repo. Choose your runner type carefully — large runners burn minutes 10× faster.
Agentic workflows can now run with GITHUB_TOKEN instead of a PAT, billing AIC directly to the organization. ULBs do not apply in this mode (no user is attributed), so control these via cost-center budgets + per-workflow caps.
permissions: copilot-requests: writeai_credits_used) — typically MAX platform engineers and cloud architects — and bump them to ~$80.