How customization supports developer engagement
by Edward Fraser Jessie Deng Eileen Thai
guidance from Steve Fadden John Chuang
This DORA Insight is an excerpt from the 2025 DORA report, State of AI-Assisted Software Development.
What we studied
As AI assistants become more common in development work, a graduate research team at UC Berkeley studied how student developers use AI-powered integrated development environments (IDEs) in practice. Using eye-tracking data and interviews, the team observed how Python developers with between one and five years of experience tackled two short tasks: one involving an unfamiliar library, and another requiring interpretation of a cryptic function.
By applying insights from this study, developers of all experience levels may find ways of working with AI coding assistants that are more attuned to their needs.
Customization as a solution
To reduce friction and better support focused work, developers and teams can customize their AI tools. Most IDEs now offer features like toggling inline suggestions, enabling “on-demand only” modes, or adjusting the style and structure of suggestions.
Repository-level config files and linked documentation can help AI assistants follow established protocols. Experimenting with these settings can align AI behavior with the cognitive demands of different tasks, helping to reduce disruption and increase the usefulness of AI assistants.
When AI gets in the way
While AI coding assistants are designed to save time and reduce effort, a study conducted at UC Berkeley found that they can also introduce friction in some tasks. For example, student developers embraced AI when handling mechanical tasks like writing boilerplate and installing packages, but when deeper understanding was needed, such as interpreting complex code, those same student developers largely ignored AI suggestions.
Eye-tracking revealed less than 1% visual attention on AI chat during interpretive tasks, compared to nearly 19% during the more mechanical ones. The students in this study often chose to complete tasks manually to retain control and comprehension, even ignoring accurate, time-saving suggestions.
Takeaway for teams
To get the most out of AI coding assistants, developers and teams should invest in customization. This study showed that AI may disrupt interpretive tasks by adding cognitive load, especially when developers are trying to make sense of unfamiliar code.
The key is aligning AI support with the nature of the task and preferences of the developer. Tuning your setup can transform AI from a source of friction into a more efficient and satisfying development experience.