AI, DORA, and the PR Bottleneck: Why Faster Code Isn’t Improving Delivery

As AI-assisted development becomes standard, many teams are seeing a familiar but confusing pattern in their DORA metrics: deployment frequency stays flat, lead time does not improve, and quality signals become harder to interpret, even as code output increases.

This session breaks down what’s happening beneath those numbers.

Rather than treating DORA metrics as the final answer, this talk shows how they act as lagging indicators in AI-driven workflows. It explains how pull request metrics provide the missing context needed to understand why delivery slows down, especially as PR volume grows and review capacity is stretched.

You’ll learn how to connect DORA and PR metrics into a single diagnostic view that reflects how modern teams actually work with AI.

Topics include:

• How AI-driven code generation affects DORA metrics indirectly

• Why DORA highlights symptoms, not root causes

• Which PR metrics surface review bottlenecks early

• How cycle time, pickup time, and rework change with AI

• How to use PR behavior to protect delivery speed and quality

This session is designed for teams who may already be using DORA metrics, but want deeper insight into how AI is reshaping their delivery system.

Note: This session is a replay from GitKon 2025.