Within occupational settings, musculoskeletal disorders are among the most prevalent and cost-intensive health issues, affecting not only workers’ health but also resulting in a reduction in their overall productivity and, consequently, economic losses to both enterprises and society. Hence there is a need to minimise their incidence in the workplace. However, creating solutions for MSD prevention is often challenging for companies. This is especially true for small and medium-sized companies which lack dedicated ergonomic expertise. On the other hand, the upsurge of generative AI has opened up new avenues to seek assistance in addressing these concerns. Nevertheless, owing to the relatively recent emergence of generative AI models, their potential applications in the workplace remain largely unexplored. To date, there seems to be a lack of studies regarding the use of generative AI in ergonomics.

Hence, this study aims to explore the use of generative AI as an ergonomic assistant for the prevention of MSDs in the workplace. Nine experts in the field of ergonomics participated in the study which included several steps and methods to follow throughout the research process to achieve this aim. First, real-world workplace scenarios were collected through a form shared with the experts. Further on, the collected scenarios were prompted to the four AI models (ChatGPT, Gemini, Claude, and a customised model called ErgoGPT), utilising prompt engineering, to which it provided answers. Finally, the answers were collected and evaluated based on experts’ feedback through a form to rate their relevance, feasibility, completeness, innovation, and accuracy. Additionally, interactive consultations were conducted with three of the experts to get a qualitative evaluation of the answers.

The study led to identifying different strengths and limitations of the AI models, and effective ways to enhance the accuracy and quality of the output. The responses generated by the AI models show a breadth of knowledge in the field of ergonomics. Additionally, the quantitative evaluation shows that ChatGPT and ErgoGPT were particularly effective, scoring high in relevance (3.1/4) and accuracy (2.8/4). However, the models’ performance varied between different criteria scoring lower in innovation and completeness, with Gemini yielding the lowest overall score of 13.8/24 due to its less innovative and comprehensive responses. The qualitative evaluation results highlighted several limitations, such as inaccuracies and incompleteness in the responses, which could limit their practical application. Additionally, the evaluation highlighted the importance of re-iteration as an approach to enhance the relevance and comprehensiveness of the AI responses by asking follow-up questions.

This study demonstrates that AI models can effectively identify potential risk factors and propose relevant and accurate ergonomic interventions. As generative AI remains an emergent field with significant potential to grow, the insights gained from this study are considered the first step towards building AI assistants for MSD prevention. Future research to further optimise AI’s role in supporting ergonomic practices, including fine-tuning for accuracy and addressing ethical considerations like data privacy and bias, will contribute to healthier and more productive workplaces.