GitHub Copilot Adds AI Adoption Cohorts to Usage Metrics
GitHub added AI adoption phase cohorts to the Copilot usage metrics API, giving enterprise and organization owners a more practical way to measure movement from code completion to agent workflows.
What changed in GitHub Copilot metrics
GitHub has added AI adoption phase cohorts to the Copilot usage metrics API. The new fields group engaged users by how they use Copilot over a rolling 28-day window, from code-first usage to agent-first and multi-agent workflows. For LinkLoot readers, the practical value is not the label itself; it is the ability to compare adoption quality rather than only counting active seats.
Key takeaways
- The new
ai_adoption_phasefield appears on user-level reports for engaged users. - Enterprise and organization reports can now include
totals_by_ai_adoption_phaseaggregates. - GitHub defines phases from code-first usage through agent-first and multi-agent usage.
- The changelog says the classification uses a two-days-in-28-days engagement rule.
- The docs still matter: access is limited to eligible organization or enterprise roles and policy settings.
Practical LinkLoot angle
Teams rolling out AI coding tools often measure license assignment, active users, or accepted completions. Those numbers miss whether developers are moving into agent workflows such as Copilot CLI, cloud agent, code review, or the GitHub Copilot app. This update gives enablement leads a cleaner workflow: segment users by adoption phase, compare pull request and merge metrics by cohort, then target training toward the phase with the biggest drop-off.
| Metric area | What it helps decide | Limitation | Source |
|---|---|---|---|
| Adoption phase | Whether users are still code-first or using agent surfaces | GitHub-specific surfaces only | GitHub Changelog |
| User-level reports | Which users need enablement or policy support | Requires the right admin access | GitHub Docs |
| Enterprise/org aggregates | Whether multi-agent cohorts correlate with delivery outcomes | Averages can hide team-level variance | GitHub Changelog |
A useful first dashboard would show total engaged users by phase, median time-to-merge by phase, and pull requests reviewed or created by phase. If agent-first users are not improving throughput, the next step is not more seats; it is workflow inspection: repository permissions, review automation, prompt templates, and where agents are allowed to act.
What to verify before you act
Check whether your organization has access to the newer Copilot usage metrics API, not only the legacy metrics endpoint. The GitHub docs note that legacy Copilot metrics endpoints were closed down on April 2, 2026 and point users toward the usage metrics endpoints. Also verify telemetry requirements, organization size thresholds, and token scopes before promising leadership a complete adoption dashboard.
For implementation ideas, pair this data with a rollout playbook from LinkLoot's AI workflow automation guide.
They are Copilot usage metrics groups that classify engaged users by product usage patterns, such as code-first, agent-first, and multi-agent workflows.