GitHub adds AI credit pools so enterprises can cap Copilot usage by cost center

GitHub has added a new spend-control option for enterprises using Copilot Business and Copilot Enterprise: AI credit pools tied to cost centers. The July 2 changelog says administrators can now cap how much of an enterprise's monthly included AI credits a cost center may draw, with the control available through the REST API today and a settings interface planned later.

The change is aimed at a practical billing problem created by pooled AI usage. GitHub says Copilot licenses contribute monthly included AI credits to an enterprise-wide pool. Without a control in place, one department or team can consume credits paid for by another before usage moves into the metered phase. The new pool limit is designed to keep each cost center inside the share funded by its assigned licenses.

How the control works

According to GitHub, administrators turn on an AI credit pool when they create or edit a cost center that contains at least one user or enterprise team. GitHub then calculates the included-credit cap automatically from the Copilot Business and Enterprise licenses assigned to that cost center, and adjusts the limit when licenses are added or removed.

The company distinguishes the new pool from a conventional cost center budget. An AI credit pool governs the included monthly credits before overage charges begin, while a cost center budget governs paid metered usage after the shared pool is exhausted. GitHub says enterprises can use both on the same cost center, giving finance and platform teams separate controls for internal allocation and external spend.

• The feature is available for Copilot Business and Copilot Enterprise on GitHub Enterprise Cloud.
• Current management is through the REST API, with a cost center settings UI still to come.
• Administrators can choose whether a cost center is blocked after reaching its included-credit cap or continues as additional spend when enterprise overages are allowed.

The update matters because AI coding assistants are moving from individual subscriptions into managed enterprise platforms. As agentic coding, CLI workflows and cloud agents increase consumption across more surfaces, organizations need clearer ways to map AI usage to teams, products and budgets. GitHub's approach does not change Copilot's underlying model behavior, but it gives enterprise administrators a tighter accounting layer for rolling out AI tools at scale.

Sources

• github.blog

Cover photo by Daniil Komov on Pexels, used under the Pexels License.