Reliability and Effectiveness of Autonomous AI Agents in Supply Chain Management
The paper examines the performance of autonomous AI agents in supply chain management through the MIT Beer Game. It identifies key factors influencing their effectiveness and highlights the agent bullwhip effect, which increases decision unreliability. A proposed reinforcement-learning framework aims to enhance reliability and reduce costs associated with these AI systems.
- ▪The study focuses on autonomous generative AI agents in multi-echelon supply chains.
- ▪Model capability is identified as the dominant factor affecting performance, with optimized models reducing costs by up to 67%.
- ▪The agent bullwhip effect amplifies decision unreliability across echelons, leading to increased decision variance.
- ▪A Group Relative Policy Optimization-based framework is proposed to improve the reliability of autonomous supply-chain agents.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Artificial Intelligence arXiv:2605.17036 (cs) [Submitted on 16 May 2026] Title:Reliability and Effectiveness of Autonomous AI Agents in Supply Chain Management Authors:Carol Xuan Long, David Simchi-Levi, Feng Zhu, Huangyuan Su, Andre P. Calmon, Flavio P. Calmon View a PDF of the paper titled Reliability and Effectiveness of Autonomous AI Agents in Supply Chain Management, by Carol Xuan Long and 5 other authors View PDF HTML (experimental) Abstract:This paper studies autonomous generative AI agents in multi-echelon supply chains using the MIT Beer Game. We identify four inference-time levers that shape performance: model selection, policies and guardrails, centralized data sharing, and prompt engineering.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
