Recovering store brand supply from disruption: AI-LLMs-enabled procurement collaboration

The adoption of Artificial Intelligence Large Language Models (AI-LLMs) in store brand management is becoming increasingly prevalent among retailers, significantly enhancing operational efficiency and decision-making. However, recent unforeseen events and tariff volatility have heightened the risks of cross-regional procurement disruptions, posing serious challenges to store brand operations. This study proposes that when overseas suppliers fail to fulfill store brand orders, retailers can leverage AI-LLMs to identify reliable domestic National Brand Manufacturers (NBMs) and initiate procurement collaborations to mitigate supply shortages. Focusing on two prevalent supply chain contracts—reselling and agency—we examine how such collaborative procurement affects supply chain participants. We find that such procurement collaboration is feasible only under a reselling contract. In this setting, the NBM can strategically leverage its wholesale price to either activate or avoid collaboration. This reflects the NBM's trade-off between preserving its national brand market share by avoiding collaboration and increasing capacity utilization while strengthening downstream relationships by activating collaboration. Interestingly, when pursuing collaboration, the NBM lowers its wholesale price, suggesting that the collaboration strategy partially alleviates the double marginalization effect. Overall, collaboration consistently benefits the retailer but harms the NBM, though total supply chain profit increases. Furthermore, we explore contractual preferences and show that both parties prefer the agency model when the commission rate is high and the NBM-sourced cost is moderate. We also conduct an extension by considering a scenario without the AI-LLMs recommendation, in which the selected NBM can only partially meet the shortage. A profit comparison further reveals that when the NBM-sourced cost is not excessively high, the supply chain under the AI-LLMs recommendation outperforms that without it.