Optimizing Supplier Selection with Large Language Models: An AI-Driven Approach for Resilient Supply Chains

Artificial Intelligence (AI) is poised to play a transformative role in Industry 4.0 and beyond. The challenges amplified by global crises, such as pandemics, have further emphasized the need for digitalization and optimization to maintain a resilient and sustainable supply chain from both buyer and supplier perspectives. Buyers must build and sustain a robust pool of suppliers capable of withstanding disruptions, while suppliers must enhance their visibility and demonstrate their capabilities effectively. Achieving this balance increasingly relies on AI-powered systems, once the appropriate digital infrastructure and data are in place. Traditional supplier selection, often based solely on cost, is no longer sufficient. Instead, procurement strategies must focus on long-term risk mitigation, requiring multi-criteria decision-making approaches. Among the AI techniques employed, Genetic Algorithms (GA) offer optimal solutions by evaluating numerous factors simultaneously. Recently, Large Language Models (LLMs) have gained attention for their ability to process vast volumes of unstructured data, enabling more comprehensive and nuanced supplier evaluations. LLMs, such as DistilGPT-2, can be integrated into decision support systems to enhance planning, forecasting, and supplier selection by considering diverse criteria, including price, delivery time, customs, taxes, production capabilities, warranty policies, and quality assurance. In this study, an AI-driven, end-to-end supplier selection approach using DistilGPT-2 was implemented. The model generated and evaluated textual data to effectively identify the most suitable supplier, demonstrating both time efficiency and decision accuracy. These findings highlight the growing potential of LLMs in optimizing supply chain management under increasingly complex market dynamics.