Large Language Models in Supply Chain Management: A SCOR-Based Assessment of Process Positioning and Implementation Stages

Large Language Models (LLMs) are attracting growing academic and industrial interest for their potential to transform supply chain management (SCM). Yet a structured assessment of where LLMs are actually applied across supply chain processes, and how mature those applications are, remains largely absent from the literature. This paper addresses that gap through a Systematic Literature Review (SLR) based on the PRISMA framework, drawing on a final corpus of 67 peer-reviewed contributions retrieved from Scopus. The Supply Chain Operations Reference (SCOR) Model is adopted as the analytical lens to classify LLM-based applications across the Plan, Source, Make, Deliver, Return, and Enable process categories, and to assess the level of their development and implementation. Results show that LLM-based applications are heavily concentrated in processes with strong informational and decision-support requirements — Enable, Deliver, and Plan collectively account for 83% of observed applications — while Make and Return processes remain largely unexplored, with Return representing a complete research gap. Regarding the development and implementation stages, the literature is predominantly exploratory: only 24% of analysed papers describe an actual operational deployment of LLMs, while prototypes and proof-of-concept applications dominate. The findings suggest that the published literature is still largely feasibility-oriented, with limited evidence of integration into enterprise information systems and industrial operations. The paper contributes a structured mapping of the current state of LLM adoption in SCM, highlights process-specific gaps, and identifies implications for both researchers and practitioners seeking to leverage LLMs in supply chain decision-making.