Modernizing vehicle inspections

A paper form made digital — and drivers treated it that way

As lead product designer for safety and innovation at Amazon Last Mile, I inherited the daily vehicle inspection — a regulated process (DVIR) governed by an international standards organization that every DSP driver completed before and after each route. The experience was a paper-inspired digital checklist: a long list of items that drivers blasted through in under 30 seconds.

Compliance rates looked fine. Trust was the problem. Fleet managers couldn’t verify whether a driver actually walked around the vehicle or just tapped through from the driver’s seat. A checked box only means something if someone actually looked.

Adding photographic evidence for fleet managers

I integrated photo capture directly into the inspection flow. When a driver flagged a defect, the camera opened with specific framing instructions for that item. The photo attached to the issue and shipped to the fleet manager — giving them visual evidence of the vehicle’s condition for the first time.

But photos only got taken when something was wrong. A clean inspection still looked identical to a skipped one.

A hardware solution enters the picture

About a year and a half into my time on the team, Amazon invested in Automated Vehicle Inspection (AVI) tunnels — camera-equipped arches installed at delivery stations that scanned vehicles as they drove through. The tunnel had to be incorporated into the existing checklist experience at stations that had one, which created a complicated UX problem: how do you fit a hardware-driven scan into a software-driven inspection without breaking either, while keeping processing time tight enough for drivers on a schedule?

My role went beyond designing the driver experience around the tunnel. I was coordinating the backend nuances — scan processing time, result syncing, failure states — with the driver-facing experience, all in service of the business need to save time and money on inspections. Drivers would sometimes jump into their manual inspection before the automated scan had finished processing, then hit a dead end. I designed a gating flow that kept the two halves in sync, and a results screen that merged automated and manual findings into a single report.

Putting AI-powered inspections in every driver’s hand

The tunnel worked — but at enormous cost, and only at stations that had one. Most didn’t. The real question: could we replace an expensive hardware scanner with a simple hand-held solution that any driver could use, at any station, using AI?

The concept: a guided camera walk-around that replaces both the checklist and the tunnel. The app directs drivers around the vehicle section by section with on-screen guides showing exactly how to frame each shot. AI validates each photo in real time. No infrastructure. No checkboxes. What cost hundreds of thousands per tunnel now runs on a phone already in every driver’s pocket.

Beyond the overlay illustrations, the app includes a real-time placement tool — simple directional prompts that update live as the driver moves, telling them exactly how to reposition for each angle. It’s intuitive by design: no instructions to read, no training to complete. The phone just tells you where to go.

The result: timestamped, geolocated photographic evidence of every vehicle angle, sent directly to the fleet portal. Consistent photo formats open the door for automated defect detection — turning every driver’s phone into an AI-powered inspection station.

Real-time guidance reduced first-time inspection errors by over 40%, with drivers consistently framing vehicles correctly on their first attempt. What used to require a training session now requires nothing — the interface teaches as you go.