Customer support · Production

Grainger uses Databricks Mosaic AI and RAG to power product search across 2.5 million products

The problem

Grainger's 2.5 million-product catalog was difficult for sales teams, call center agents, and non-specialist customers to navigate, while over 400,000 daily product updates made real-time accuracy hard to maintain with their existing search infrastructure.

Workflow diagram · grounded in source
1
Customer or agent submits product query
trigger
“when they call in and want to make a purchase, they can't necessarily look up the specifications on our website, which adds to the workflows of our sales and customer service teams”
2
ETL pipeline and vectorization
integration
“The Databricks Data Intelligence Platform facilitated an efficient workflow that streamlined the entire data management process and minimized errors — from data extraction and cleaning to transformation and loading to vectorization”
3
Vector index real-time sync
integration
“Vector Search automating the synchronization of product data from the source to the search index, Grainger can support high volumes of product embeddings and real-time queries”
4
RAG retrieval for agent query
ai_action
“They could now retrieve relevant results for call center agents even with the diverse range of customer queries the company receives daily”
5
LLM response generation via Model Serving
ai_action
“Databricks Model Serving, a unified interface for managing multiple large language models (LLMs), enabled Grainger to easily switch between different LLMs and query them through a single API”
6
Accurate product response delivered
output
“customer inquiries are met with precise, contextually appropriate responses”
Reported outcome

Grainger achieved significant advancements in search recall and discoverability, with sales teams and call center agents equipped with faster and more accurate product retrieval, saving time and reducing errors.

Reported metrics
Search recall and discoverabilitysignificant advancements in search recall and discoverability
Product retrieval speed and accuracyfaster and more accurate product retrieval capabilities
Employee time savedsaves time
Query response latencyaccurate and near-instantaneous results
Reported stack
DatabricksDatabricks Mosaic AIVector SearchModel Serving
Source
https://www.databricks.com/customers/grainger
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

Grainger achieved significant advancements in search recall and discoverability, with sales teams and call center agents equipped with faster and more accurate product retrieval, saving time and reducing errors.

What tools did this team use?

Databricks, Databricks Mosaic AI, Vector Search, Model Serving.

What results were reported?

Search recall and discoverability: significant advancements in search recall and discoverability; Product retrieval speed and accuracy: faster and more accurate product retrieval capabilities; Employee time saved: saves time; Query response latency: accurate and near-instantaneous results (source-reported, not independently verified).

How is this customer support AI workflow structured?

Customer or agent submits product query → ETL pipeline and vectorization → Vector index real-time sync → RAG retrieval for agent query → LLM response generation via Model Serving → Accurate product response delivered.

Related customer support cases