ConciseSignal
Following

AI speeds up code writing but slows code review

A recent Harvard study across multiple software firms finds that while AI coding agents boost code production by 20-30%, they do not lead to more finished software. The time taken for human code review increases by around 49% after the introduction of AI tools, with more reviewers and comments per pull request. Most companies now use some form of AI agents, but humans still do most code review work.

Why it mattersFor firms adopting AI in software development, expected productivity gains may be offset by slower human review. This suggests that AI's impact on the software delivery pipeline may be limited unless review processes evolve.

Sources covering this

Ars TechnicaAI coding agents generate more code, but not more software7:43 PM →TechRadarAI agents are great at writing new code — but then the human bottleneck slows everything down3:20 PM →
Concise Signal DailyEnterprise AI, security & business tech.Weekdays, 7am Eastern · Sample issue

More in AI