Paper Type
Complete
Paper Number
PACIS2026-1792
Description
Organizations are increasingly integrating large language models (LLMs) into their supply chain management systems to enhance decision making under uncertainty. A particularly critical application area is supply chain resilience, where LLMs can help firms interpret and respond to disruption-induced volatility. However, organizations lack prescriptive guidance on building systems that can systematically detect disruptions using statistical methods and manage the resulting volatility in demand and inventory planning. To address this shortcoming, our action design research develops a resilience analytics system that detects disruptions using deviation-based thresholds and employs an LLM to adjust the noisy demand through shock-cut and half-life decay mechanisms and thereby stabilize inventory levels throughout the disruption period. By doing so, we propose design principles to guide the development and deployment of LLM-augmented resilience systems in supply chain environments.
Recommended Citation
Tripathi, Pritesh; Sharma, Hardik; Mishra, Shivansh; Kumar, Bagesh; Herath, Savindu; and Shrestha, Yash Raj, "Designing an LLM-Augmented Resilience Analyst for Disruption-Driven Demand Adjustment" (2026). PACIS 2026 Proceedings. 5.
https://aisel.aisnet.org/pacis2026/practioner/practioner/5
Designing an LLM-Augmented Resilience Analyst for Disruption-Driven Demand Adjustment
Organizations are increasingly integrating large language models (LLMs) into their supply chain management systems to enhance decision making under uncertainty. A particularly critical application area is supply chain resilience, where LLMs can help firms interpret and respond to disruption-induced volatility. However, organizations lack prescriptive guidance on building systems that can systematically detect disruptions using statistical methods and manage the resulting volatility in demand and inventory planning. To address this shortcoming, our action design research develops a resilience analytics system that detects disruptions using deviation-based thresholds and employs an LLM to adjust the noisy demand through shock-cut and half-life decay mechanisms and thereby stabilize inventory levels throughout the disruption period. By doing so, we propose design principles to guide the development and deployment of LLM-augmented resilience systems in supply chain environments.
Comments
16-Practitioner