GitHub Copilot Usage API Adds AI Adoption Cohorts

What changed

GitHub's Copilot usage metrics API has been enhanced with the introduction of AI adoption phases, enabling a more nuanced understanding of how users engage with Copilot. Previously, metrics focused on general activity; now, each active user is categorized into one of four distinct phases based on their usage patterns over a rolling 28-day period. These phases are: Phase 0 (No cohort), for users not meeting engagement criteria; Phase 1 (Code first), for those engaging with code completion and/or IDE agent mode; Phase 2 (Agent first), for users interacting with a single GitHub-based agent surface like Copilot cloud agent, code review, or CLI; and Phase 3 (Multi-agent), for users engaging with two or more GitHub-based agent surfaces or the new GitHub Copilot app.

A new ai_adoption_phase field is now available on user-level reports. For enterprise- and organization-level reports, a totals_by_ai_adoption_phase array provides aggregated metrics broken down by these phases. These aggregated metrics include total engaged users, average user-initiated interactions, average code generation and acceptance activity, average lines of code added and deleted, average pull requests created, merged, and reviewed, and average median time-to-merge. Importantly, these aggregated metrics report the average per user within each phase, rather than a simple sum. Each ai_adoption_phase value includes a version field, starting at v1, to accommodate future evolution of Copilot's product surfaces without disrupting historical data context.

Why it matters for builders

This enhancement provides developers and their organizations with a clearer picture of AI tool adoption maturity. Instead of just knowing if a developer is using Copilot, administrators can now see *how* they are using it—whether they are primarily focused on code completion, exploring specific agent functionalities, or leveraging multiple AI-powered workflows. This granular data can inform targeted training programs, documentation improvements, and rollout strategies, helping teams maximize their investment in AI tools.

Practical impact

Enterprise administrators and organization owners with access to Copilot usage metrics via the REST API can now leverage these new insights. By analyzing the ai_adoption_phase data, they can identify trends in developer behavior, such as users progressing from basic code completion to more advanced agent-based workflows. This allows for proactive support and resource allocation, focusing enablement efforts on phases where developers might need additional guidance or where opportunities for deeper integration exist. The API's ability to track cohort progression over time enables organizations to measure the effectiveness of their AI adoption initiatives and demonstrate the evolving value of Copilot within their development teams. This release can also be combined with the teams filter for even more detailed analysis.

Caveats and source limits

The AI adoption phase metrics are exclusively available to enterprise administrators and organization owners who already have access to Copilot usage metrics through the REST API. The source does not provide details on how these phases are calculated beyond the described engagement criteria, nor does it offer independent benchmarks or comparative data against other AI coding assistants. The classification logic is subject to evolution, as indicated by the versioning of the ai_adoption_phase field, meaning that future updates to Copilot's surfaces could alter how users are categorized.