Overview
- A Harvard-led study published Monday found that after firms adopted AI coding agents lines of code rose about 30 percent, commits rose about 20 percent, and pull requests rose about 23 percent without a matching rise in completed tracked features.
- The research shows review work slowed sharply: average review time lengthened from 7.03 days to 10.48 days, the share of PRs with changes requested nearly doubled, and comments per PR increased roughly 35–49 percent depending on the metric.
- Most firms use AI to help with reviews but AI accounts for only a minority of review activity, producing about 23.3 percent of review comments and directly affecting about 10.8 percent of pull requests.
- The study found about a 14 percent rise in the share of employees doing code reviews but did not detect clear net employment losses attributable to AI after cross-checking firm analytics and LinkedIn data.
- A key caveat is the paper’s March 2026 data cutoff, so the findings may not reflect tool or process changes since then and suggest firms must redesign review workflows to turn AI-written code into finished product.