To main content

Generative AI and Developer Workflows: How GitHub Copilot and ChatGPT Influence Solo and Pair Programming

Abstract

This study investigates the impact of generative AI (GAI) tools, like ChatGPT and GitHub Copilot, on software development. GAI tools can automate repetitive tasks, provide code suggestions, and enhance problem-solving. We conducted a case study in a large-scale agile organization, where we examined how GAI tools influence development efficiency and developers’ workflow. During 25 days of observations, we observed 49 programming sessions. GAI was used in all 16 solo programming sessions and 30 of 33 pair programming sessions. Additionally, we conducted 14 interviews with developers and team leaders. In solo programming, GAI tools were said to improve efficiency and reduce stress, allowing more time for solving complex tasks. In contrast, the impact of GAI in pair programming was mixed. GAI tools were less frequently used in pair programming compared to solo programming. Our findings also highlight changes in software development, with developers leveraging GAI for information retrieval and coding assistance. However, challenges such as dependency on AI tools and the accuracy of AI-generated suggestions persist.

Category

Academic chapter/article/Conference paper

Client

  • Research Council of Norway (RCN) / 321477
  • Research Council of Norway (RCN) / 309344

Language

English

Author(s)

Affiliation

  • University of Oslo
  • SINTEF Digital / Software Engineering, Safety and Security
  • Unknown

Year

2025

Publisher

University of Hawaii at Manoa

Book

58th Hawaii International Conference on System Sciences (HICSS 2025)

Issue

58

ISBN

978-0-9981331-8-8

Page(s)

7379 - 7390

View this publication at Cristin