GitHub expands Copilot analytics to nine usage scenarios

Engineering leaders often view Copilot adoption through an overly narrow lens: how many licenses have been assigned, how many suggestions have been accepted, and how many lines of code have been generated. That picture excluded people who regularly discuss tasks with Copilot Chat or launch an agent workflow in VS Code. The work was happening, but the dashboard captured only part of it.

GitHub has added the two missing activity sources to the Copilot impact dashboard: Copilot Chat and the Agents panel in VS Code. The chart now covers nine usage scenarios. The same metrics are available in 28-day enterprise and organization reports through the chat and vscode_agent fields.

What this looks like in practice

Consider a team of 20 developers. Five of them barely use autocomplete but regularly open chat, investigate errors, and ask the agent to modify several files. Previously, these employees could have been mistaken for passive Copilot users. Their activity is now visible separately.

GitHub has also renamed the Passive users cohort to Chat-first users. The new name is more accurate: these are not necessarily people who ignore the tool. They are active in Copilot but did not use, on at least two days during the 28-day period under review, the feature required to qualify for one of the adoption stages.

The engagement rule is simple: a user is considered active in a particular scenario if they used it on at least two separate days during a 28-day period. One person can appear in several metrics at once. For example, a developer using agent mode within Copilot Chat will be counted in chat statistics, while work performed through the separate Agents panel will count toward the VS Code Agent metric.

What not to confuse

Adding the new data does not change Copilot's adoption-stage model. Chat and the Agents panel appear as feature activity, but they do not, by themselves, move a user to the next stage. Cohort sizes have not changed either, and the API still returns the Chat-first group as No Cohort.

This is an important limitation. Growth in the Chat line cannot automatically be interpreted as greater team maturity or proof that development has accelerated. The dashboard shows where employees use Copilot, but it does not say whether they save time, ship changes faster, or produce fewer defects.

There are technical constraints as well. The new fields are available only in 28-day reports at the organization and enterprise levels. Historical reports generated before the update retain the previous seven scenarios. Access is restricted to owners, billing managers, or members whose role permits viewing Copilot metrics, and the metrics collection policy must be enabled.

What engineering leaders should do

First, check whether the share of “passive” users was caused by incomplete telemetry. Then compare chat and vscode_agent with clear engineering outcomes: pull request cycle time, task duration, the number of requests for rework, and post-release defects.

A useful experiment is to select two similar teams and, for one month, compare changes in their workflows rather than the number of Copilot interactions. The new analytics makes it easier to distinguish how the tool is being used. But decisions about renewing licenses should still be based on engineering outcomes, not an attractive rise in activity on a chart.