- AI hasn't shifted the bottleneck from coding to code review, and coding was never the bottleneck.
- Many teams have multiple approved changes waiting in batches after code review, indicating the bottleneck is downstream (e.g., testing, deployment).
- Industry studies show AI increases coding speed and merge rates, but review times and PR sizes also rise.
- Accumulated unreleased changes in batches increase risk; speeding up code review just moves pressure to the real bottleneck.
- The real constraint may be manual steps, cumbersome approval processes, or deployment issues, not coding or reviewing.
- Research often stops at code merge, missing post-review delays, which hides the true bottleneck.
- AI initiatives succeed or fail based on addressing batch problems; fixing code review alone won't improve value delivery.