Bridging theory and practice in AI-driven supply chains: Prioritizing LLM adoption challenges and SCOR-based applications

The integration of artificial intelligence (AI) into supply chains is reshaping decision-making, yet the role of large language models (LLMs) remains under-theorized and weakly connected to operational frameworks such as the Supply Chain Operations Reference (SCOR) model. Unlike conventional AI tools, LLMs are generative, rely on unstructured data, and introduce interpretability risks, creating distinct opportunities and challenges for supply chain management (SCM). This study adopts a two-phase, theory-driven approach to examine these challenges. First, LLM adoption barriers were identified through a comprehensive literature review and structured using the Technology–Organization–Environment framework and Dynamic Capability Theory, then validated using a Delphi study with industry experts. Second, the Combinative Distance-based Assessment and Comparative Solution (COCOSO) method was applied to prioritize the identified challenges. Results show that “Understanding LLM Outputs,” “Risk of Over-Reliance on AI,” and “Limitations in Complex Optimization Problems” are the most critical barriers, indicating that interpretability and human–AI interaction outweigh data-related concerns emphasized in prior AI research. In contrast to existing AI-focused SCM research, this study contributes by providing a theory-informed, expert-validated prioritization of LLM-specific adoption challenges and illustrating how these challenges relate to value creation across SCOR process layers. The study further offers a SCOR-aligned decision-support roadmap for managers and policymakers and establishes a structured foundation for future empirical research on LLM-enabled supply chains.