GitHub announced on November 6, 2026 that the GitHub Copilot impact dashboard now includes engagement from Copilot Chat and the VS Code Agents window. The same two feature entries are available in eligible enterprise and organization 28-day usage-metrics reports. For QA and engineering leaders, this improves visibility—but it also makes metric definitions and validation more important.
What changed in the GitHub Copilot impact dashboard
GitHub says the agent adoption chart now covers nine Copilot experiences. It also renamed the Passive users cohort to Chat-first users. That label applies to people who are active in Copilot but have not used a qualifying feature on at least two days in the 28-day period, so they do not qualify for an adoption phase.
The new report entries are chat and vscode_agent in copilot_feature_engagement.totals_by_feature. GitHub is explicit about an important boundary: Chat and the VS Code Agents window are engagement signals only. They do not change adoption phases or cohort sizes. Chat also includes captured Copilot Chat usage such as agent mode, so an agent edit can appear in the Chat signal.
Why this matters for QA engineers
Usage is not validation. A rising Chat or VS Code Agent count can show where teams are experimenting, but it does not prove that generated changes passed tests, reduced defects, or met review standards. QA teams should use the dashboard as a discovery signal, then join it with independent delivery evidence: pull-request outcomes, CI status, test coverage changes, escaped-defect trends, and human review findings.
A practical QA checklist for the new signals
- Confirm the reporting window. Use a fixed 28-day period and record the extraction date. GitHub notes that periods before this update include only the original seven features.
- Test the event definition. Verify that a user needs activity on at least two days to be considered engaged, and document which teams, licenses, and client versions are in scope.
- Prevent double-counting. A person can be counted in more than one feature. Do not add per-feature totals and call the result unique users without a deduplication method.
- Separate engagement from adoption. Keep Chat and VS Code Agent usage out of any calculation intended to reproduce GitHub’s adoption phases. Compare the API’s No Cohort value carefully with the dashboard’s Chat-first label.
- Reconcile with quality evidence. Sample agent-assisted pull requests and compare CI pass rate, rerun rate, review rework, and post-release defects with a matched non-agent baseline.
Example measurement plan
Start with one repository and one release train. Export the dashboard/API counts weekly, identify users with VS Code Agent engagement, and select a comparable control group. For every sampled pull request, capture whether Copilot was used, the test command and result, reviewer-requested changes, and any defect linked after release. Review the results after four weeks; this keeps an adoption dashboard from becoming an untested productivity claim.
Access and limits to document
According to GitHub, the data is available in enterprise and organization 28-day reports to authorized owners, billing managers, or roles with the View Copilot Metrics permission, and the usage-metrics policy must be enabled. Treat access configuration as part of the test plan. A missing signal may indicate permissions or policy configuration rather than lack of user activity.