AI coding agents generate 30% more code but no more software
AI coding agents generate more code, but not more software
A study of over 700 firms found that AI coding agents boost code output by 30% and pull requests by 23%, yet don't increase completed software features. The gains are absorbed by a human review bottleneck: review times balloon 49%, change requests nearly double, and comments per PR rise 35%. Firms see no significant employment changes.
Any efficiency increased during the actual coding phase is absorbed by downstream constraints in the production process.