Enterprise Development Teams Achieve 39% Higher PR Merge Rate with Cursor AI Agent

Title

Enterprise Development Teams Achieve 39% Higher PR Merge Rate with Cursor AI Agent

Content

A University of Chicago study published in late 2025, analyzing tens of thousands of Cursor users across more than 1,000 organizations and 10,000 developers, found that companies experience a 39% increase in merged pull requests after Cursor's AI agent became the default coding environment. The research compared organizations that had been using Cursor before the agent was released against a baseline group that had not, measuring the rate of merged PRs relative to time trends. The study found no significant change in the PR revert rate and a slight decrease in bugfix rates, suggesting the productivity gains did not come at the expense of code quality. Senior developers were found to be more effective with AI agents than junior developers — a counterintuitive finding. For every standard deviation increase in years of experience, the rate of accepting AI-generated code edits increased by approximately 6% relative to the mean. The researchers theorized that experienced developers are better at using custom rules and managing context effectively, and are more confident in evaluating AI-written code changes. Analysis of conversation-starting requests showed that 61% of requests were for code implementation, with the remainder split between explaining code/errors and planning actions. Senior developers were notably more likely to engage in planning before generating code. The study measured the economic impact of AI coding tools on software engineering organizations and found that the productivity gains were statistically significant and sustained over the measurement period. Cursor CEO Michael Truell noted at Fortune Brainstorm AI that the findings surprised the team, as the conventional assumption had been that junior developers would benefit most from AI coding assistants.