A practical way to measure which AI coding tools your team actually uses

Cursor versus Copilot market-share charts can disagree because they count different things. Shattered.io lays out a practical audit for measuring use inside your own team, combining survey answers, licence data and a short staff poll rather than pretending one chart settles the argument.

Watch Desk analysis

What happened

The guide proposes a roughly 70-minute audit using a public developer survey, GitHub Copilot seat data, a three-question pulse survey and an app inventory from device management. A small Python script counts survey responses and reports Wilson confidence intervals, which help show when a small sample makes an apparent lead uncertain.

It also separates four measures the kit can calculate directly: tools used in the past 30 days, the primary assistant, assigned seats and active seats. Paid seats, weekly active users and AI-assisted code share need vendor-console data, which the guide leaves for a separate ledger.

Why it matters

Each data source sees a different slice. Surveys are broad but self-reported; seat records show licences, not necessarily whether someone opened the tool; telemetry can show activity, but vendors do not share it with rivals. Put them side by side and label their denominators, and a team can make a more honest local comparison than by borrowing a vendor’s reach figures.

The guide also flags a timely wrinkle: it says Copilot licence arrangements changed on 28 October 2026, so an assigned seat may not mean a developer actively used the assistant. That distinction matters when a manager is deciding what to renew, not merely what looks good in a chart.

Our read

This is useful because it turns “which tool is winning?” into a question a team can actually answer about itself. The catch is that it measures your organisation, not the whole market, and some of the most revealing usage figures require access to vendor telemetry. The sensible takeaway is to compare several imperfect measures, not crown a universal champion from one convenient number.

What to watch

• Whether the survey sample reflects the team being measured.
• Whether seat assignments are mistaken for active use.
• Which measures require vendor-console access, and who can obtain it.

Discussion spark: If you had to choose just one measure for deciding which coding assistant to keep, would you trust reported usage, assigned seats or active-use telemetry most?

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.