The reason for this discrepancy can be found directly in the code review process, which on average takes much longer after the introduction of AI coding agents. Overall, the average “review process” time between submitting a pull request and merging it into the codebase increases by 49 percent on average after AI agents are introduced. That effect can also be seen in more granular data, as “the proportion of pull requests with requested changes nearly doubles.”[ing]and the number of comments per pull request increases[ing] by 35%” after the change of AI agent, the researchers write.
In response to this change, researchers found a 14 percent increase in the proportion of workers performing code reviews after the introduction of AI agents. They also write that they “cannot attribute significant changes in employment to AI” after looking at total active workers at Jellyfish and cross-referencing those companies’ LinkedIn data.
Pull requests need changes much more frequently in the “agent coding” era.
Pull requests need changes much more frequently in the “agent coding” era.
Credit: Chen and Stratton
While AI could theoretically also help with this review process, researchers found that, so far, that impact has been marginal. Although 80 percent of companies measured used some form of AI code review as of March 2026, AI agents were only responsible for 23.3 percent of all review comments and 10.8 percent of all pull requests, suggesting that humans were still responsible for the vast majority of this work.
AI agents are still a relatively new part of the coding world, of course, and there have been major updates and improvements to their output even since this study’s data cutoff in March 2026. And while 95 percent of companies in the study have deployed AI coding agents at this point, many are no doubt still going through a learning process about when and how to best deploy them. These types of “coding time vs. review time” tradeoffs could improve as software engineering teams gain more experience with the pros and cons of using an AI agent on particular coding problems.
For now, however, letting AI write your code seems like a double-edged sword, as increases in coding speed are offset by similar increases in human code review time and effort. It’s the kind of result that makes us wonder if the considerable time and expense to get AI coding agents up and running is really worth it for most companies.
