AI Agents for Inventory Control: Human-LLM-OR Complementarity

A new study on arXiv shows AI agents combining LLMs with traditional operations research (OR) algorithms outperform either method alone in inventory control. Testing over 1,000 scenarios with real and synthetic data, the hybrid approach handled demand shifts and seasonality better. In a human-in-the-loop experiment, human-AI teams also achieved higher average profits than humans or AI working in isolation, challenging prior findings that such collaboration degrades performance.