OBJECTIVES: Excel-based Health Economic (HE) models are widely used in Health Technology Assessment (HTA) due to their flexibility and user-friendliness. However, these models often face challenges, especially with complex calculations, debugging issues, and heavy formula usage. Although there is a transition from Excel to advanced software like R due to R's advanced capabilities, the shift is slow because many users are unfamiliar with R. This study explores whether Gen-AI based agent plugins can serve as a savior by maintaining the Excel platform while enhancing its functionality. Specifically, this research investigates if Gen-AI plugins can assist in reviewing formulas, providing interpretations, and suggesting optimizations within Excel.

METHODS: A proof-of-concept Gen-AI plugin was developed and integrated into Excel. The plugin's functionality includes extracting formula from cells, creating a prompt using the extracted formula, and sending this prompt to a Gen-AI model. The AI model processes the prompt and returns a response and displays within Excel plugin window. The Gen-AI plugin's capabilities were tested to see if it could assist in reading complex formulas, providing more efficient formula suggestions, interpreting formulas in layman’s terms, and generating simple visual aids to trace dependencies and precedents.

RESULTS: The Gen-AI plugin successfully read complex Excel formulas and provided suggestions for more efficient alternatives by following programming best practices. It interpreted complex formulas into simple text, making them easier to understand. Additionally, the plugin generated layman graphical representations of trace dependencies and precedents, helping users grasp the context quickly. The AI also identified potential errors and provided suggestions to correct inconsistent flows, significantly improving the model's reliability and user comprehension within a limited time.

CONCLUSIONS: This proof-of-concept study demonstrates that Gen-AI plugins can significantly enhance Excel-based HE models' functionality. This paves way to embrace AI enhancements to assist model implementation and verification to unlock higher levels of precision and efficiency.