Thread · 01 · Walkaround
The walkaround, unfucked.
UX critique, flow diagrams, confidence architecture, and operator/customer interventions for VUMO's AI-assisted vehicle walkaround. Working notes in the open.
Working model · The current flow
Five steps the walkaround goes through.
First-pass decomposition. Each step is a decision surface for UX, copy, and human intervention. Real output lands here as threads close.
1
Initiate
Operator (porter, service advisor, customer self-serve) opens the walkaround on device. Vehicle ID, context, stakes, expected output are all established before the first photo.
2
Capture
Guided photo capture. The interface must coach, not demand. Confidence of capture shown in real time. Retry and re-capture are first-class — not buried.
3
Detect
AI damage detection runs. Every finding has visible confidence + human intervention affordance. No silent assertions. Nothing passes to the next stage without clear state.
4
Confirm
Operator (and/or customer) reviews findings. Agree · disagree · add context. This stage is where trust is built or destroyed — the UI must make agreement as easy as correction.
5
Report
Output is a report a human can read, share, and reference. Language, structure, and follow-up actions all matter. The report is the product the customer remembers.
Principle
Every AI output has a confidence signal, a correction path, and a clear next action. No silent assertions.
Principle
The interface coaches the operator. It never shames or blames. Humans trust tools that make them look competent.