Automotive After-Sales Intelligence

After-sales service lives or dies on accuracy: a misdiagnosis at intake cascades into the wrong labor codes, the wrong parts, and a vehicle that comes back for the same fault — driving up warranty cost and eroding customer trust. The root causes were spread across the service chain: imprecise fault diagnosis, slow access to technical information, and inefficient dealer-to-HQ communication.

We proposed and architected four AI modules that target those failure points directly. A Smart Customer Portal simplifies booking and gives an early, plain-language read on the likely fault. A Service-Bay Diagnostic Assistant sharply raises accuracy at intake and in the bay by standardizing customer complaints and blocking inappropriate labor entries. A Workflow Agent automates dealer-to-HQ communication. And a Unified Technical Knowledge Base turns a passive document archive into a conversational, multimodal “thinking brain.”

The whole platform is designed to improve itself. A continuous feedback loop (RLHF and self-reflection) is tuned on the strongest possible signal — whether a vehicle returns for the same repair — alongside intake-acceptance, chatbot-helpfulness, and customer-satisfaction signals, so every interaction makes the next diagnosis more accurate.

It is built as a microservices platform with a Persian/English OCR ingestion pipeline, hybrid semantic-plus-keyword search over hundreds of technical documents, and grounded answers that always cite their source document and page — refusing to answer rather than hallucinate when the knowledge base comes up empty.