Retail agents improve answers by linking data from many sources
Toward Agentic Retrieval-Augmented Generation for Enterprise Retail Workflows over Heterogeneous Data Sources
Artificial IntelligenceInformation Retrieval
Summary
Answering questions in retail is hard because information is spread over different types of data like product details, orders, and customer reviews. The authors created a new test called RetailBridge that mimics these complex questions by connecting clues across these different sources. They also built a system called RetailRAG that helps language models figure out which data to use and how to combine it step-by-step. Their tests show that the hardest part is reliably connecting information from multiple sources, not just making the models bigger.
What this means in practice
- •For retail data engineers: Improve automated systems that answer detailed retail queries by reliably linking different data types into coherent responses.
- •For customer service developers: Build advanced agents that handle complex customer questions pulling evidence from orders, products, and reviews all at once.
Authors
Jeonghwan Choi, Jibin Hwang, Hayoon Park, Hyeonjae Cheon, Hwanjun Song
Abstract
Enterprise retail workflows require agents to chain evidence across heterogeneous sources, from structured product and order records to unstructured inquiries and reviews, yet existing agentic RAG benchmarks rarely capture this challenge. To enable controlled evaluation of such workflows, we introduce RetailBridge, a compositional retail benchmark built from persona-driven simulations. Its questions are constructed by linking entities and intermediate answers across heterogeneous sources, requiring agents to identify the relevant sources, carry evidence between them, and produce a grounded answer. To systematically evaluate LLMs on these retail workflows, we develop RetailRAG, a retail-aware framework equipped with auto-schema briefing, source-routed plan chaining, and diagnostic plan revision. RetailRAG consistently outperforms existing agentic RAG baselines across all evaluation axes, serving as a strong reference system for assessing LLM agents on retail workflows. On RetailBridge, evaluating ten LLM backbones with RetailRAG reveals multi-source chaining as the dominant bottleneck across providers and model scales, suggesting that reliable retail agents require explicit cross-source orchestration rather than scaling alone. Code and data are publicly available at \url{https://github.com/DISL-Lab/RetailBridge}.