Towards Autonomous and Reliable Supply Chains

This project explores how Generative AI can transform supply chain management from rule-based automation to true autonomy. Building on the MIT autonomous supply chain testbed, it integrates multiple AI agents that learn, adapt, and coordinate decisions across forecasting, inventory, and replenishment—cutting costs by up to 40%. The research addresses a critical challenge: ensuring reliability in large language model agents, which often exhibit unpredictable behaviour. A novel meta-reinforcement learning framework and policy reconciliation techniques are proposed to stabilise performance, enabling trustworthy, data-efficient decision-making in complex, safety-critical environments. The work aims to establish a foundation for autonomous, reliable supply chains across diverse domains.

This research is conducted via The Agentic AI Research and Innovation (AARI) Initiative which focuses on the next frontier of agentic systems through Grand Challenges with the academic community and Microsoft Research.