github's new copilot dashboard tells you which users are still just autocomplete people

GitHub shipped a new Copilot impact dashboard for enterprise administrators and organization owners on July 22, 2026. The dashboard extends Copilot usage metrics with a view that groups developers into AI adoption phases and shows the measurable output differences between them.

three phases, one diagnostic

the dashboard carves your licensed users into four buckets:

• phase 1 — code-first: developers using copilot primarily for completions, inline suggestions, and chat for code
• phase 2 — agent-first: developers using copilot agent mode for longer autonomous tasks
• phase 3 — multi-agent or copilot app: developers running orchestrated agent sessions or the standalone Copilot app
• passive: licensed but not engaged — the cohort most enterprise admins do not want to think about

each cohort card surfaces pull requests merged per user per month, median PR merge velocity, user count, share of overall users, and average lines of code per day per user. github also adds an "adoption multiplier" — a throughput and speed comparison between the passive cohort and the average of engaged users, meant to put a number on the "what are we missing?" question.

what the dashboard is actually measuring

this is github's answer to the "we bought seats but no one changed how they ship" problem. the metrics are output-facing — PRs, merge velocity, lines per day — rather than just activity signals like "user opened copilot three times this week."

the six-month trend charts and per-cohort recommended next steps are the interesting operational layer. admins get specific guidance on moving a phase-1 developer into phase-2 usage, not just a chart showing adoption is flat.

the honest caveat

lines of code per day and PR count are noisy proxies for developer productivity. a phase-3 multi-agent user who merges one careful refactor per week might be delivering more value than a phase-1 user who merges twelve small autocomplete-assisted fixes. use the dashboard to find unlicensed-but-active users, identify who has never touched agent mode, and spot teams that went backwards after an initial spike. do not use it to rank individuals by throughput.

the dashboard is available to enterprise administrators and organization owners who already have access to Copilot usage metrics. it builds on the ai_adoption_phases field that landed in the metrics API earlier this year.

for teams that are just setting up Copilot workflows, start with the usage basics first — check your AI credits before this dashboard has anything meaningful to phase-sort. the credit meter we covered earlier this week is the prerequisite view.