When to use large language models for digital manufacturing in supply chains?

The emergence of large language models (LLMs) offers opportunities to digitally transform manufacturing and supply chain operations by improving decision-making efficiency and automating manufacturing processes. In this paper, we quantitatively explore the value of LLMs and uncover LLM adoption decisions in the supply chain. We formulate a stylized model including one supplier and one manufacturer to examine models with traditional manual operations and with LLM deployment. Under two models, we analytically compare the operational and economic outcomes to capture when to adopt LLMs. We identify the specific conditions under which LLMs benefit the supplier and the manufacturer as well as consumers. Moreover, the supplier and manufacturer share an identical threshold condition for benefiting from LLM adoption. In addition, we extend the basic models by exploring other economic factors influencing LLM adoption decisions, i.e., LLM license, LLM deployment under a cost-sharing mechanism, government sponsorship and the fixed selling price. The finding reveals that LLM customization facilitates LLM adoption by better aligning LLM capabilities with a firm’s specific operational context; an optimal proportion exists to maximize the likelihood of LLM adoption under a cost-sharing mechanism by appropriately distributing the deployment costs; the government sponsorship significantly reduces the obstacle of LLM adoption. Finally, we offer useful managerial implications to guide the LLM deployment in the supply chain.