AI coding agents generate more code, but not more software
Summary
A study by Harvard researchers found that using AI coding assistants and agents helps software teams produce more lines of code, but it does not improve the overall completion of software projects. The reason is longer and more demanding code review processes, which slow down merging AI-generated code into final products.Key Facts
- AI coding agents increase the amount of code produced by about 30%.
- Commits (saved code changes) rise by 20%, and pull requests (requests to add code) increase by 23% after AI use.
- Despite more code, the completion rate of software features does not improve significantly.
- Code reviews take about 49% longer and involve more requests for changes after AI agents are introduced.
- The number of comments on pull requests goes up by 35% after AI tools are adopted.
- More workers (14% increase) perform code reviews to handle the extra workload.
- The research used data from over 700 software companies involving more than 700,000 employees from 2021 to early 2026.
- The study found no significant change in employment levels due to AI coding tools.
Read the Full Article
This is a fact-based summary from The Actual News. Click below to read the complete story directly from the original source.
