AI impact

See how pull requests assisted by different AI coding tools — or no AI tools — perform across metrics like throughput, cycle time, batch size, and review time.

The AI tools → AI impact page helps you understand how pull requests assisted by different AI coding tools — or no AI tools — perform across metrics like throughput, cycle time, batch size, and review time.

This makes it easier to answer questions like: Is AI-assisted work moving through your system faster or slower? Which teams benefit most from AI tools? Are AI-assisted pull requests staying small enough to review well?

How to use AI impact

Use AI impact to understand how AI-assisted pull requests move through your delivery system:

• Speed: See how pull request cycle time and throughput change depending on AI use. Compare teams with different adoption levels, and pay attention to the share of review time versus time in progress.

• Batch size: Monitor pull request batch sizes to keep review quality high. AI assistants make it easy to generate large pull requests, and large changes are harder to review well.

Use these metrics to understand where AI helps, where it introduces new bottlenecks, and what your teams should look into next.

Available metrics

• PRs merged (throughput)

• Cycle time

• Time to first review

• Time in review

• Batch size

Available selections

Comparison type

First, use the tabs to select how the pull requests are grouped:

• AI tools: GitHub Copilot vs Claude Code vs Cursor vs everything else.

• Modes: local changes (editor/CLI) vs cloud agent vs review agent vs everything else.

• AI used: AI-assisted PRs (based on the tools and modes you select) vs everything else.

The structure of the page is the same for all comparisons.

Included tools

Select which AI tools (GitHub Copilot, Claude Code, and Cursor) are included in the comparison.

Included modes

Select which modes (local changes (editor/CLI), cloud agent, and review agent) are included in the comparison.

There's also an option to include only high-confidence matches in the local changes mode.

Frequently asked questions

Why is the total under "PRs merged" lower than the sum of its parts?

A single pull request can be associated with multiple AI tools and modes. For example, a developer might use both Cursor and Claude Code on the same PR. In that case, the PR is counted once in the Cursor category and once in the Claude Code category.

How do you determine which AI tools to show for a PR?

Read more about our automatic AI tool detection.