Multi-Agent AI Automates Retail Price Taxonomy.

For global retailers managing millions of active items, maintaining consistent pricing and implementing an Every Day Low Price strategy is crucial but manually unfeasible. Inconsistent pricing across product variants can negatively impact customer perception and cannibalize sales. To tackle this challenge, a new scalable, context-aware Multi-Agent Framework has been developed to automate the creation of "Lines and Ladders" pricing taxonomies. This framework employs specialized Large Language Model (LLM) agents that work collaboratively. These agents are designed to identify key product attributes, extract multi-modal values from various data sources, and apply hierarchical grouping logic to construct coherent pricing structures. Evaluated on real-world enterprise data and already deployed in production, the 3-Agent system achieved an F1-score of 0.83 for "Lines," significantly outperforming single-agent baselines by reducing cognitive overload. The system demonstrates over 90% precision and 75% recall in Food & Consumables, and 80.2% assignment accuracy in the less structured General Merchandise catalog.