Erik Malmgren, Isak Määttä:
Exploring the Integration of Generative AI in Pair Programming and Code Review,
summary, report, June 2025.

Abstract:

Software development teams rely on collaborative practices such as pair programming and code review to ensure quality, share knowledge, and maintain productivity. While effective, these practices are time-consuming and often involve repetitive or low-level tasks that divert attention from higher-order problem solving. The rise of generative AI (GAI) tools such as GitHub Copilot introduces opportunities to streamline collaboration, yet their role in real-world workflows remains underexplored. Existing research offers limited guidance on how GAI can complement, rather than disrupt, team dynamics, especially in industrial settings where domain knowledge is essential.

This thesis investigates how GAI can be effectively integrated into pair programming and code review within E.ON’s development teams. A mixed-methods approach was applied, beginning with a literature review, developer interviews, and an observational study to inspire and inform two experimental iterations. GitHub Copilot was chosen as the GAI tool, supported by structured prompt guides tailored to specific tasks.

Results show that in pair programming, GAI enhanced productivity in code generation, refactoring, and documentation, while also supporting complex problem decomposition through conversational workflows. In code review, GAI effectively surfaced syntactic and stylistic issues, enabling human reviewers to focus on architectural concerns. Findings suggest GAI adds greatest value as a complement to human judgment, emphasizing the need for structured prompting and thoughtful workflow integration.