A self-hosted private LLM giving dealership teams instant, secure answers from their own data
Keeping Dealership Data Private with a Self-Hosted LLM for Sales and Service Teams
Key results from this project
Dealership Data Kept on Private Infrastructure
Customer Data Sent to Public AI Providers
Project Summary
A multi-rooftop automotive dealer group approached Hatzs wanting the speed of generative AI without sending customer records, deal structures, or pricing data to a public AI service. Their sales, service, and BDC staff were losing time searching across the DMS, CRM, inventory feeds, OEM bulletins, and internal policy documents to answer routine questions. Hatzs designed and deployed a private large language model hosted entirely inside the dealer group's own cloud environment, connected to their internal systems through retrieval-augmented generation, so every answer is grounded in the dealership's own data and stays under its control.
High performance components for demanding applications
They needed a technology partner to support its digital strategy
Data privacy was the non-negotiable starting point. Customer PII, credit applications, and deal margins could not be processed by third-party AI providers. Hatzs built the platform around an open-weight model running on private infrastructure, with role-based access that mirrors the dealership's existing permissions, so a salesperson, a service advisor, and a finance manager each see only what their role allows.
- Full Data Privacy: An LLM hosted on the dealer group's own infrastructure, with no customer, deal, or pricing data sent to public AI APIs.
- Grounded Answers from Dealership Data: Responses drawn from the DMS, CRM, live inventory, service history, OEM bulletins, and internal policy documents, with sources shown.
- Role-Based Access Control: Permissions aligned with existing dealership roles so each user only retrieves data they are authorized to see.
Our Approach & Solution Delivery
Key implementation aspects included:
Self-Hosted Private LLM
An open-weight language model deployed inside the dealer group's private cloud, keeping every prompt and response within the dealership's own environment.
Dealership Knowledge Retrieval (RAG)
A retrieval layer connected to the DMS, CRM, inventory feeds, and document libraries, so answers are grounded in current dealership data with cited sources.
Role-Aware Department Assistants
Sales, service, BDC, and F&I assistants that respect existing permissions, giving each team fast answers without exposing data outside their role.
Results & Outcomes
The dealer group now gives its sales, service, BDC, and F&I teams instant, source-backed answers from their own systems while keeping every piece of customer and deal data inside the dealership's private environment.
- Full Data Privacy: An LLM hosted on the dealer group's own infrastructure, with no customer, deal, or pricing data sent to public AI APIs.
- Grounded Answers from Dealership Data: Responses drawn from the DMS, CRM, live inventory, service history, OEM bulletins, and internal policy documents, with sources shown.
- Role-Based Access Control: Permissions aligned with existing dealership roles so each user only retrieves data they are authorized to see.
- Department-Specific Assistants: Tailored assistants for sales, service, BDC, and F&I teams working from the same secure knowledge base.
- Audit and Compliance Readiness: Full logging of prompts, retrieved sources, and responses to support privacy and consumer-finance compliance reviews.
