How Does Generative AI Usage Affect the Coding Performance of Developers?

Generative AI (GenAI) techniques have recently advanced at a rapid rate. However, the impact of GenAI on developers’ coding performance remains an empirical question. On the one hand, GenAI tools such as GitHub Copilot could provide coding suggestions for developers at a fast speed, enabling them to use the suggested codes directly, which could speed up their coding process. On the other hand, developers need to verify the accuracy of the suggested codes, leading to additional cognitive load, which could hurt their coding performance. Collaborating with a mid-size information technology global organization, we apply difference-in-differences (DID) analysis to over 27 weeks of proprietary data. Our findings reveal that using GenAI tools leads to an increase in the number of user stories completed per week among developers. We also explore the moderating effect of developers’ working experience to test the mechanism. The implications are both theoretical and practical.