An Experience Report on Leveraging LLMs for GUI Generation: Automating Coding to Prioritise Creativity

Giovanna Broccia, Alessandro Borselli, Maria Rosaria Cefaloni, Franco Delcorno, Alessio Ferrari

GUI Design, LLMs, ChatGPT, DeepSeek, Requirement Engineering 

The design of graphical user interfaces (GUIs) is a complex and time-consuming process that begins with identifying user roles and gathering requirements through interviews, surveys, or workshops. Designers then create low-fidelity sketches or digital wireframes, organising information into logical sections and selecting visual elements to enhance usability. This iterative process often demands extensive refinement based on stakeholder feedback, making mockup creation—especially for interactive prototypes—a time-consuming task. In particular, the mockup development process often entails spending significant effort on clerical activities, such as programming and debugging tasks, rather than concentrating on creativity, human interaction, and quick feedback cycles with stakeholders. This paper investigates whether large language models (LLMs) can assist GUI designers in streamlining the design process—reducing time and effort while maintaining design quality—enabling them to focus on the human aspects of creativity and user interaction by offloading technical programming tasks to the machine. We document our experience designing a dashboard for predictive maintenance in railways, illustrating how LLMs can support key tasks such as requirement analysis, information organisation, and mockup generation and refinement. We discuss insights and lessons learned, including the importance of clear requirements, the impact of LLM choice, and the benefits of iterative refinement in achieving stakeholder alignment. Our study shows that LLMs can support the GUI design process by automating specific tasks, thereby reducing design effort and enhancing the overall quality and satisfaction of the final product.