Grainger builds a GenAI RAG-powered product search to help call center agents navigate 2.5 million products

Overview

Grainger implemented a retrieval augmented generation (RAG) search system on the Databricks Data + AI Platform, using Databricks AI Search and Model Serving to let customer service and sales teams retrieve accurate product information from a 2.5 million-item catalog with over 400,000 daily product changes, improving search recall and reducing errors for call center agents.

The challenge

Grainger's customer service team struggled to rapidly retrieve accurate product information from a 2.5 million-product catalog to support its high-traffic e-commerce site. The search system needed to understand a buyer's role and industry application, interpret incomplete or non-technical queries from non-specialist customers, and keep pace with over 400,000 product updates occurring daily.

The solution

Grainger built a retrieval augmented generation (RAG) search system on the Databricks Data + AI Platform. Databricks AI Search provided the data engineering and management foundation for RAG, while Databricks Model Serving gave a unified interface to switch between and query multiple LLMs through a single API, with automated data vectorization and embedding synchronization keeping product indices accurate.

Reported business value

Grainger saw significant advancements in search recall and discoverability across its 2.5 million products, with faster and more accurate product retrieval that saves time, reduces errors and lets sales and call center staff assist customers more efficiently, while handling over 400,000 product changes daily.