Transforming Software Development with Generative AI: Empirical Insights on Collaboration and Workflow
Rasmus Ulfsnes, Nils Brede Moe, Viktoria Stray, Marianne Skarpen
TL;DR
The paper investigates how Generative AI (GenAI) transforms software development workflows and team dynamics through an empirical study of 13 practitioners. Using an exploratory multi-case design, it captures diverse GenAI usages (e.g., ChatGPT, Copilot) across roles and highlights benefits such as accelerated learning, higher productivity, and increased motivation, alongside challenges like data confidentiality and integration gaps. A key finding is that GenAI can disrupt traditional knowledge sharing and learning loops in agile teams, while also enabling new forms of collaboration such as pair prompt engineering. The work demonstrates practical implications for integrating GenAI into daily software work, suggesting both efficiency gains and the need for disciplined governance to preserve team learning and coordination benefits.
Abstract
Generative AI (GenAI) has fundamentally changed how knowledge workers, such as software developers, solve tasks and collaborate to build software products. Introducing innovative tools like ChatGPT and Copilot has created new opportunities to assist and augment software developers across various problems. We conducted an empirical study involving interviews with 13 data scientists, managers, developers, designers, and frontend developers to investigate the usage of GenAI. Our study reveals that ChatGPT signifies a paradigm shift in the workflow of software developers. The technology empowers developers by enabling them to work more efficiently, speed up the learning process, and increase motivation by reducing tedious and repetitive tasks. Moreover, our results indicate a change in teamwork collaboration due to software engineers using GenAI for help instead of asking co-workers which impacts the learning loop in agile teams.