GitHub has added per-user AI credit consumption to the Copilot usage metrics API. The new ai_credits_used field shows how many AI credits each user consumed, using the same credit-consumption data that feeds GitHub’s usage-based billing API.
The field is available in single-day and 28-day user-level reports at the enterprise and organization levels. It is a small changelog item, but it answers a growing admin question: once AI coding tools are billed by usage, who is consuming the credits?
That makes Copilot governance more concrete.
Usage-based billing needs usage visibility
GitHub moved Copilot plans to AI Credits on June 1. That shift makes cost less like a fixed seat count and more like a pool of model and feature consumption. Admins need to understand how credits move across teams before a bill arrives or a budget conversation turns vague.
The new field helps connect adoption and consumption. An organization can see whether credit usage is concentrated in one team, spread across a broad developer base, or rising in a way that maps to a product launch, migration, or heavy agent use.
GitHub says the metric is meant for analyzing consumption, not as a billed total. That distinction matters. Finance teams should still rely on billing data for invoices. Engineering and platform teams can use the usage metrics API to understand behavior.
The field is broad by design
GitHub says ai_credits_used is an overall per-user total. It is not currently broken down by feature, model, or surface.
That limitation reduces how precise the metric can be. A user may consume credits through chat, inline edits, agent mode, code review, CLI, or other Copilot surfaces. The API field will show the total, but it will not explain which workflow drove it.
Even so, broad per-user visibility is useful. It gives admins a first-order signal for where to investigate. If one team has unusually high consumption, the next step is not necessarily restriction. It may be training, policy, budget allocation, or a closer look at whether the team is using agent workflows that need guardrails.
Cost governance is becoming product governance
The deeper shift is that AI coding platforms now have an operating layer. Enterprises need to manage seats, permissions, models, data policy, usage, code review, and spend. Copilot credit metrics sit inside that layer.
This also changes how value is argued. A developer tool can no longer be justified only by activation or seat adoption. Teams will increasingly ask what credits produced: pull requests, tests, reviews, migrations, support fixes, or reduced cycle time.
GitHub’s new field does not answer the value question by itself. It supplies the consumption side of the equation. Teams still need to pair it with engineering outcomes and code quality signals.
The next ask is feature-level cost
The next useful metric would be a breakdown by surface, feature, or model. Admins do not just need to know that a user spent credits. They need to know whether the credits went into high-value agent tasks, casual chat, expensive model choices, or automated workflows that should be scoped.
For now, ai_credits_used is the right starting point. It turns Copilot’s usage-based billing from an abstract plan change into data that enterprise admins can query.
