This podcast discusses a paper which investigates the impact of generative AI on the productivity of software developers. The authors conducted three randomised controlled trials at Microsoft, Accenture, and a Fortune 100 company to evaluate the impact of GitHub Copilot, an AI-based coding assistant. They found that across all three experiments, software developers who used Copilot experienced a statistically significant 26.08% increase in completed tasks, along with substantial increases in code updates and compilations. The study also highlights the heterogeneous nature of these productivity gains, with less experienced developers demonstrating higher adoption rates and experiencing larger productivity increases.