AI pull request overview
The AI pull request overview report shows your existing PR metrics—cycle time, size, revert rate, and more—broken down by the percentage of AI-generated code in each PR. Use it to understand how AI-authored code moves through your development process differently than human-written code.
This report is powered by AI Code Insights. AI Code Insights must be deployed to developer machines before data will appear in this report.
When to use AI pull request overview
This report helps teams:
- Compare SDLC metrics across AI code levels — See whether PRs with high AI code percentages move faster, get more review comments, or revert more often than PRs with little or no AI code.
- Spot quality and velocity signals — Identify patterns like longer cycle times or higher revert rates for AI-heavy PRs that may point to areas where AI tools need better guidance or oversight.
- Inform AI tool strategy with data — Use concrete metrics to decide where to invest in AI tooling, where to adjust review practices, and how to set expectations for AI-assisted development.
What the report shows
The report is a dashboard of PR metrics, each displayed as a time series chart broken down by AI code percentage band. Each chart shows how that metric differs across levels of AI code contribution.
| Metric | What it measures |
|---|---|
| PRs opened | Number of PRs opened per period |
| PRs merged | Number of PRs merged per period |
| PR cycle time | Time from PR open to merge |
| PR size | Lines of code changed per PR |
| Review comment density | Number of review comments per PR |
| Review turnaround | Time from review request to first review |
| PR revert rate | Percentage of PRs that were reverted |
| Allocation | Percentage of PRs categorized as bug fixes, features, or maintenance |
Each metric can be drilled down to the individual PR level.
How AI code percentage is calculated
Each PR is assigned an AI code percentage based on the code changes in its commits:
AI additions ÷ (AI additions + human additions) × 100
Where:
- AI additions is the total number of lines added that were classified as AI-authored across all commits in the PR
- Human additions is the total number of lines added that were classified as human-authored across all commits in the PR
This percentage is recalculated periodically as new commits are pushed. PRs with no additions (e.g., delete-only PRs) have no AI code percentage and appear in the “Unknown” band.
Understanding AI code percentage bands
Each PR is grouped into one of five bands based on its AI code percentage. Charts in the report show a separate series for each band.
| Band | Criteria | Description |
|---|---|---|
| 0% AI | pct_ai = 0 |
Entirely human-written — no AI-authored lines detected |
| <33% AI | 0 < pct_ai < 33 |
Mostly human-written with some AI assistance |
| 33–66% AI | 33 ≤ pct_ai ≤ 66 |
Roughly equal mix of human and AI-authored code |
| >66% AI | pct_ai > 66 |
Predominantly AI-authored code |
| Unknown | pct_ai is null |
AI code percentage could not be determined (e.g., no commit data yet) |
Use the AI code % filter to focus the report on specific bands—for example, comparing only “0% AI” against “>66% AI” to see the starkest contrasts.