Screenshot: GitHub Changelog GitHub Copilot Gets AI Adoption Phase Tracking, CLI Fixes Free User Model Picker
GitHub added a four-phase AI adoption classification to the Copilot usage metrics API, letting enterprise admins see how deeply their teams are using agent features. A separate CLI release the same day restored model selection for Free and Student plan users.
GitHub shipped two Copilot updates on May 29 that affect how enterprise admins measure usage and how free-tier users interact with the model picker.
AI adoption phases in the usage metrics API
The Copilot usage metrics API now classifies each engaged user into one of four phases based on their Copilot activity over a rolling 28-day window. The classification requires at least two days of product use within that window to count.
The four phases:
- Phase 0 (No cohort): Didn’t meet the engagement criteria for any phase.
- Phase 1 (Code first): Used code completion or IDE agent mode.
- Phase 2 (Agent first): Used a single GitHub-based agent surface, meaning the Copilot cloud agent, code review, or Copilot CLI.
- Phase 3 (Multi-agent): Used two or more GitHub-based agent surfaces, or the GitHub Copilot App.
The progression is intentional. Phase 1 captures developers who have adopted AI completions. Phase 2 marks the jump to agentic workflows. Phase 3 reflects users who have integrated multiple agent surfaces into their regular work.
A new ai_adoption_phase field appears on user-level reports. Enterprise and organization-level reports get a totals_by_ai_adoption_phase array for per-phase breakdown. The field also includes a version key, starting at v1, so the classification logic can evolve without breaking historical queries.
This is primarily useful for enterprise admins trying to answer “are teams actually using agent features, or just code completion?” before drawing conclusions from aggregate token or seat counts.
Copilot CLI v1.0.56: model picker restored for Free and Student users
The same day, GitHub released Copilot CLI v1.0.56 to fix a regression from v1.0.55, which had locked Free and Student plan users to Auto-only model selection. The regression was a one-day turnaround: v1.0.55 shipped May 28, v1.0.56 shipped May 29.
Free and Student users can now select models other than Auto in the model picker, as intended. The release also fixed a layout issue where the ThemePicker wasn’t fitting correctly in 120-column terminals.
Some context on the timing
The adoption phase feature lands the week before Copilot’s Usage-Based Billing switches on June 1. The two aren’t directly related, but the combination gives enterprise admins a clearer picture of which users have moved into agentic workflows and are therefore more likely to consume premium request credits under the new billing model.
The CLI model picker fix is more straightforward: locking free-tier users to Auto was apparently unintentional, and GitHub corrected it within 24 hours.