Situation
Valcon is proud to be working with one of the leading global automotive OEMs, leveraging Generative AI to assist technicians across their European dealer network in finding the right information from a myriad of technical manuals, to ensure that they can repair customers’ vehicles correctly and in a timely manner.
As modern vehicles have become increasingly complex, there is a vast and ever-growing library of technical manuals required to support each model release, as well as urgent bulletins and campaigns. The result is a time-consuming, manual and specialist-intensive process in identifying the correct information and procedures to repair a customer’s vehicle, and get it right first time. This challenge is multiplied by technicians across 53 European countries needing this information in their native language.
The business wanted a smart, scalable solution that would allow technicians to access information rapidly without compromising on quality and traceability.
Approach
The solution is a Retrieval-Augmented Generation (RAG) Chatbot that allows 14,000+ technicians across 3,000 garage workshops to query the complete data source of technical manuals via a natural language interface. Building the solution on the customer’s Snowflake data platform was the logical choice, allowing all the data preparation to be completed (Chunking, Embedding, Vectorization) using Snowflake DB and Cortex, and native LLMs with language support. Utilising a broad library of AWS services the solution is secure, robust and scalable to support up to 2,000 simultaneous users.
With a strong evaluation and test strategy, we ensure that the answers provided by the Chatbot are and remain transparent, accurate and compliant with the EU AI Act. We also put in place appropriate data and insights to support the evaluation of usage, performance, sentiment and success for our customer and their technicians.
Results
- The first deployed version of the Chatbot is focused on supporting experienced technicians with common vehicle problems that represent 80% of work undertaken within the garages. It allows the technicians to search based on “symptom” style questions – for example, “the steering wheel is shaking” and be provided with answers that have been strongly filtered to ensure only relevant information is provided.
- The Chatbot is also able to handle general, local and brand-specific ‘jargon’ that might be used by technicians to ensure responses are as effective as possible and time is not wasted.
- The next phase of development is to expand the platform to support a greater number of vehicle models that the Chatbot will have access to data for, and to increase the number of languages (currently English, French and Italian) to allow for an enterprise-wide roll-out
