Walmart Deploys Multi-Agent LLM for Inventory Forecasting

Walmart Global Tech senior engineer Aditya Mulik presented a sequential multi-agent LLM workflow that adds real-time context to backward-looking ML forecasts for fresh-food inventory across 4,700 U.S. stores. Four agents gather metadata, synthesize sales data, ground predictions with weather and event APIs, and output natural-language directives like increasing apple stock 15%, with store managers reviewing each recommendation.