The introduction of ChatGPT by OpenAI and its versatile use cases in software development led to the creation of new tools supporting developers or engineers in their daily tasks. Many well-known software companies have launched their versions of AI assist-ed coding tools. Notably, GitHub Copilot, examined alongside ChatGPT in this thesis, stands out among these innovations. This study evaluates whether access to AI assisted coding tools has a noticeable impact on the productivity of software developers working within a large-scale agile framework.

To this end, more than two years of selected system-based metrics from La Mobilière insurance’s agile organisation and 125 questionnaire responses from the company’s GitHub Copilot users were analysed.

The findings suggest that without system-based metrics that accurately isolate the coding period, too many variables could influence the outcome. However, considering the survey responses from professional software developers at La Mobilière, there is an indication of a significant productivity increase in the daily tasks experienced by developers from various backgrounds. Additionally, the data from La Mobilière offers a positive economic outlook for organisations contemplating the implementation of GitHub Copilot in their software development teams on a larger scale. Developers benefit from integration into their preferred Integrated Development Environment (IDE), which reduces the need to switch contexts to other tools. Furthermore, GitHub Copilot users enjoy a chat interface within the IDE, providing suggestions that can be seamlessly incorporated into their existing code.