Original title: Adapting to Evolving Requirements: Agentic AI for Retail Supply Chain Operations
Lei Zheng, Liping Yang, Zihao Li, Guodong Lyu +2 · September 2026
Retail supply chains depend on connected decision tools that have to change when business needs shift. Large language models, or LLMsLarge language models that process and generate natural language, like GPT or Qwen., can help users update these systems using plain language. However, one change can take many paths through a multi-step system, creating different results. Researchers designed a graph-guided system of AI agents to solve this. In their system, specialized domain agents outline allowed changes, while a central coordinator searches through limited intervention routes. The system then tests and compares candidates using supply chain performance metrics. Working with a large retail partner, the team tested 100 warehouse needs gathered from workers. Across GPT, Qwen, and DeepSeek, this method increased end-to-end success from 72--76% up to 79--83%, while also improving correctness over direct LLM updates.
When planning a road trip with friends, each person decides what to pack based on the weather. If the forecast changes, everyone might need to adjust, but their adjustments affect each other—like if one person brings a tent, another might skip a sleeping bag to save space. Normally, people pack separately, which can lead to overlaps or gaps. In this paper, the system acts like a trip coordinator who tests different packing strategies across the group, using a structured way to find the best overall plan that minimizes problems when requirements change.
Retail supply chain operators must handle many connected decisions that change over time, often leading to errors. This framework helps them find the best intervention paths, reducing mistakes and improving overall success rates in real-world operations.
The abstract does not address scalability to other supply chain domains, long-term adaptation effects, or real-time implementation. Evaluation is limited to warehouse requirements from one retail partner and specific language models.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
Adapting to Evolving Requirements: Agentic AI for Retail Supply Chain Operations — Lei Zheng, Liping Yang, Zihao Li et al.
Abstract licence: CC0
Explanation generated by Synthesis from the abstract — it is not the original paper. This is a preprint, so it hasn't necessarily been peer reviewed.