Quality sampling is a recurring manual check against silent quality erosion: review a random subset of merged AI-generated changes for defects that automated checks do not measure. section 9.4.3
A practical sampling loop:
- Review ten randomly selected AI-generated changes each week, focusing on naming, architectural alignment, unnecessary complexity, and missed test edge cases rather than bugs already caught by CI. section 9.4.3
- Track recurring findings alongside first-attempt success, developer feedback, and other health signals; sampling supplies evidence that aggregate metrics may miss. section 9.4.3
- Reduce sampling frequency only after stable results justify lower monitoring overhead; one team moved from weekly to biweekly after three stable months. section 9.4.4
A regression automated metrics missed
A model update changed TypeScript generic handling, producing technically correct but unnecessarily complex annotations. Sampling exposed the accumulated complexity even though CI reported no bugs. section 9.4.3
Figure 9.7 Visualization of the potential dashboard - state of the art.

Use the dashboard visualization to situate quality sampling among the broader health-check signals and monitoring decisions. section 9.4.4
