From User Needs to Smart Ecosystems Through LLMs: The Smartifier Case Study

Smartifier is a conversational platform that leverages Large Language Models (LLMs) to help non-technical users design, implement, and manage smart objects. It addresses persistent challenges in the Internet of Things (IoT) domain, where the “making community” and End-User Development approaches have yet to empower novices fully. The system features two distinct chat interfaces: one dedicated to creating customized devices, and another for generating automation rules. Building on an iterative design model, Smartifier translates high-level, daily-life needs into hardware selections and automatically compiled software, ensuring correct APIs for seamless integration. A technical evaluation compared different ChatGPT-based models and temperature values, revealing an optimal configuration that curbs hallucinations and maximizes reliability. In a user study with 17 participants lacking ICT expertise, Smartifier achieved high correctness (averaging over 91% across all tasks) and high scores regarding usability and user satisfaction. A thematic analysis of conversations revealed insights around user need expression, iterative negotiation, and system feedback. These findings will inform future refinements, especially regarding user experience design and the handling of advanced functionalities in complex real-world settings. While the study’s sample size is limited, findings confirm that Smartifier effectively bridges complex technical tasks and everyday needs, reducing cognitive load and streamlining the creation of IoT solutions. These results highlight the potential of LLM-driven platforms to simplify and democratize technology development, suggesting new directions for future research on EUD and software generation for non-expert users.