Supply chain mapping through retrieval-augmented generation: applications to the electronics industry

Abstract

This paper presents a novel methodology for automated multi-tier supply chain mapping, leveraging Retrieval-Augmented Generation (RAG) and network science techniques. We developed an RAG-based approach that extracts supplier-customer relationships from unstructured public data sources, including SEC 10-K filings and earnings calls. The extracted entities are structured into a directed supply chain graph and analysed using network science metrics such as centrality, modularity, and path length. The case study focuses on three of the largest contract manufacturers in the electronics industry: Hon Hai Precision Industry (Foxconn), Flex Ltd., and Jabil Inc. Our findings demonstrate that Generative AI (GAI), specifically LLMs enhanced with RAG, can construct scalable and comprehensive supply chain graphs. The proof of concept is successful, as evidenced by the construction of a directed supply chain graph encompassing 4,644 nodes and 8,341 edges, covering three of the largest contract manufacturers in the electronics industry.