AI support

AI-guided before/after comparison — live on site.

The guided capture process instantly compares every fresh image with the last confirmed image of the same vehicle area from the record, flags deviations and lets the employee decide right away — with separate Confidence and Risk scoring.

How the AI works

It proposes — the human decides.

Guided Capture

Enforce standard views (bumpers, doors, rims …) → images are comparable.

Load history

Last confirmed image of the area from the vehicle record.

Pre-filter + AI

A cheap on-device diff flags suspicion → only that goes to the vision model (cost control, no unnecessary upload).

Change Event

Area · type · severity · Confidence + Risk — with both images as evidence.

Decision on site

Confirm / Unsure / No damage → becomes part of the Evidence Chain and a training label.

AI Change Event · Before/After

Rear bumper right · new scratch

Pre-inspection 07.08. 08:14 → current 08.08. 17:42

Before

After

AI Confidence94 %

Risk score78 / 100 · HIGH

Confidence ≠ Risk

Two values, clear triage.

Confidence = how certain the AI is that something has changed.
Risk = economic/liability significance (severity × position × cost × responsibility period).

Auto-accept

Confidence low · Risk low

Minor/known deviation — logged, no action needed.

On-site check

Confidence medium · Risk medium

Employee is prompted for a targeted close-up & confirmation.

Escalation

Confidence high · Risk high

Mandatory detail photo, attribution to the responsibility period, case for case management.

EU-compliant

Assistance with mandatory confirmation, EU hosting/on-device pre-filter, AI provenance in the audit trail.

Cost control

Only suspicious areas go to the model; paid comparisons only after sign-off, with a hard limit.

Data asset

Every confirmation/rejection becomes a training label — detection gets better and cheaper over time.