Built with
AI assistant for Suzuki owners answering spare parts questions, delivered in collaboration with Zen Digitix.
The problem is specific and genuinely annoying. A car owner needs a part, does not know the exact designation, is not sure whether the one they found fits their model, and has no easy way to sanity-check a price. The existing options are calling a workshop or trusting a marketplace listing. Both are slow and neither is transparent.
The app puts that knowledge in a conversational interface. Users ask in natural language, describing the problem rather than naming the component, and get back part information, compatibility guidance, and price estimates. Coverage spans popular models including the Bolan, across everything from brake pads to engine oil filters.
The engineering challenge in a domain assistant is not the model, it is grounding. A general language model will answer a parts compatibility question confidently and sometimes wrongly, and in this domain a confident wrong answer costs the user money and a return trip. The value sits in constraining responses to the actual catalogue rather than in the fluency of the reply.
Built with React Native and TypeScript for cross-platform delivery. Mobile is the correct surface here because the moment of need is usually standing next to the car or in a parts shop, not at a desk.
The collaboration with Zen Digitix paired the mobile and AI engineering with real automotive domain knowledge, which is the part that cannot be substituted with a better prompt.
Private engagement, case study available on request.
More work