NuVIN AI
Designing trust into an AI-powered vehicle inspection platform.
Overview
NuVIN AI uses computer vision to automate vehicle condition assessments for dealers and insurers. The UX challenge wasn’t the AI — it was trust. Dealers didn’t believe the scores, insurers needed audit trails, and the product had to serve both simultaneously.
Problem
Early prototypes showed high accuracy but low adoption. Dealers overrode AI scores 70% of the time in beta, not because the scores were wrong — they just didn’t understand how they were generated.
Process
Conducted a two-week contextual inquiry at dealerships to map the mental model dealers had around vehicle assessment. The insight: dealers trusted their own physical process, so the interface needed to show its work, not just its answer.
Key design decisions
- Explainability-first score breakdown (not just a number — a reason)
- Photo evidence linked directly to flagged line items
- Dispute flow designed to be fast (<2 min) to reduce friction without incentivizing frivolous disputes
Outcome
After redesign, dealer override rate dropped from 70% to 28%. Dispute resolution time cut in half. The component library built for this project was adopted across three internal product lines.