The AI Return dashboard combines several distinct signals about AI tool adoption and the impact of AI-assisted work on delivery outcomes. This article is the entry point — it explains what each metric category measures and how to read them together.
Metric categories
Adoption — who is using AI tools, how often, and which tools.
AI-assisted PRs
AI-assisted Commits
Distinct AI users per week (covered in AI User Engagement Segments)
AI User Engagement Segments — Power, Casual, Idle, New
Volume — how much code is flowing through AI-assisted work.
AI Lines of Code
Avg Lines per PR (AI-assisted vs non-AI) — covered in AI Lines of Code
Impact — does AI usage correlate with measurable delivery improvements?
AI Lead Time Comparison (with Coding, Waiting for Review, In Review, Ready to Deploy sub-stages)
AI Intensity — and its scatter plots against Cycle Time, Throughput, and Bug %
Investment — what AI usage costs and how seats are utilized.
AI Tool Cost
Data sources
AI Return derives signals from connected AI tool integrations and overlays them on existing Git activity:
GitHub Copilot — real seat data. Cost comes from the seat count and the plan: free $0, individual $10, team $19, enterprise $39 per seat.
Claude Code — activity data only. Seats are estimated from unique active users, priced at $25 per seat by default.
Cursor — same estimate, priced at $20 per seat by default.
Gemini — ingested and shown in the AI usage views, but it does not feed AI Tool Cost or the engagement segments.
Two things follow from that, and they matter before anyone quotes a cost-per-pull-request figure:
The Claude Code and Cursor prices are defaults, not your contract. They can be overridden per request, so check what your organization has set before treating the total as real spend.
A connected tool can contribute usage but $0 of cost. That makes cost-per-PR look better than it is. Read adoption and investment side by side rather than one without the other.
If a tool isn't connected, its panel on this dashboard stays empty.
How to read AI Return
Start with Adoption. Are AI tools being used at all? By whom?
Look at Volume. Is AI-assisted work a small share of total code or a meaningful one?
Check Impact. Are AI-enabled teams actually moving faster than peers?
Round out with Investment. Are the costs justified by the impact you observed, and are the seats you pay for actually utilized?
Caveats
Correlation, not causation. AI Intensity scatter plots show patterns, not proof.
The "AI-assisted" label on a PR or commit means at least one commit was authored on a day the developer used an AI tool — not that the AI wrote the code.
Engagement segments (Power, Casual, Idle, New) are documented in AI User Engagement Segments.
Related articles
AI Return dashboard
AI Lead Time Comparison
AI-assisted PRs
AI-assisted Commits
AI Lines of Code
AI Intensity
AI Tool Cost
AI User Engagement Segments