AI Coding Agents Speed Up Development—but Increase Code Review Time by 49%

AI coding agents may help developers write code faster, but new research suggests that productivity gains can come with a significant trade-off: more time spent reviewing and revising AI-generated code.

AI coding agents make pull requests take longer to merge

The discrepancy can be seen directly in the code review process. On average, the time between submitting a pull request and merging it into the codebase increased by 49% after companies deployed AI coding agents.

The impact was also visible in more detailed measures of review activity. The share of pull requests that requested changes nearly doubled, while the number of comments per pull request increased by 35% after companies adopted AI agents, according to the researchers.

More developers are involved in code reviews

Companies also increased the number of workers performing code reviews. Researchers found that the percentage of workers involved in reviewing code rose by 14% after AI coding agents were introduced.

However, the researchers found no evidence that the change was caused by reductions in employment. After examining the total active workforce across Jellyfish and cross-referencing employment data from LinkedIn, they wrote that “any significant change in employment cannot be attributed to AI.”

AI code review has not replaced human reviewers

AI tools could theoretically help manage the additional review workload, but their impact has been modest so far. By March 2026, 80% of the companies measured were using some form of AI code review. Even so, AI agents were responsible for only 23.3% of all review comments and 10.8% of all pull requests.

Those figures suggest that human engineers still perform most code review work, even in organizations that have adopted AI coding agents.

The AI coding productivity trade-off

AI coding agents remain a relatively new part of software development, and their output has undergone significant updates since the study’s March 2026 data cutoff. Although 95% of the companies surveyed had deployed AI coding agents by that point, many teams are still learning when and how to use them most effectively.

The trade-off between coding time and review time may improve as software engineering teams gain more experience directing AI agents toward specific coding problems. For now, however, the research suggests that faster code generation can create a comparable increase in human review effort.

That makes AI coding agents a double-edged tool for software teams. Any increase in coding speed may be offset by the additional time and expense required to review, revise, and merge AI-generated code.