GitLab research finds AI coding gains do not automatically speed delivery

GitLab research covered by InfoQ finds that AI tools can accelerate coding without necessarily improving overall software delivery. The bottlenecks often move to review, security, governance and coordination.

That distinction is important for companies measuring AI productivity. Faster code generation does not automatically translate into faster releases if downstream processes are unchanged.

The finding supports a more realistic view of AI developer tools: they can improve parts of the workflow, but organizations still need process changes to capture end-to-end gains.