Motor claims assessment · by aQb Solutions

AI-Assisted Vehicle Damage and Repair Estimate Assessment.

AI Motor Claims Assessment is AI-assisted vehicle damage and repair estimate assessment that recommends a decision for a person to confirm.

Video transcript

Property and casualty insurance fraud costs the United States an estimated 45 billion dollars a year, according to the Coalition Against Insurance Fraud. AI Car Insurance Claims by aQb Solutions assesses a motor own-damage claim from the policy, the registration certificate and up to three accident photos. It checks that the policy is valid, then marks each dent, crack and scratch with a severity. Every garage estimate line is compared with the visible damage, so lines the photos do not support are not paid. It then calculates the payable amount after the excess. And it recommends auto-approval, human review or not eligible, with reasons. Serious structural damage and mismatches always go to a person, who makes the final decision. The documents shown are samples. Assess faster. Pay what the photos support. Keep a person in control.

AI Motor Claims Assessment by aQb Solutions (shown in the film as AI Car Insurance Claims) helps motor insurers assess own-damage claims: it marks damage on accident photos, checks the garage estimate line by line and recommends a decision for an assessor to confirm. It checks the policy and vehicle registration document, marks visible damage on up to three accident photos, compares the garage estimate line by line and calculates a preliminary amount after the excess. It recommends one of three outcomes, with reasons: recommended for straight-through approval, confirmed by a person; human review; or not eligible. A person makes the final decision and authorises any settlement.

Impact you can measure

Illustrative targets, to be confirmed in a pilot on your own data.

MinutesTarget time to a first assessment
100%Claims checked against policy, vehicle and documents
60%+Target reduction in manual claim handling
SecondsTo mark and classify damage on the photos
100%Repair estimate lines checked against the photos
InstantPreliminary claim amount, with the reasons shown
100%Serious structural damage routed to a person
24/7Claim intake

What it does

Reads the claim papers

Extracts the customer, policy number, validity dates, vehicle number, coverage and excess from the policy, the owner and model from the vehicle registration document, and each line of the garage estimate. Amounts stay in the currency of the document.

Validates the claim

Checks that the policy is valid on the accident date, the vehicle matches across the policy and the vehicle registration document, the cover includes own-damage accidents, the documents are complete and damage is visible. A claim is stopped if the photos show different cars or the papers do not belong together.

Finds the damage on the car

Inspects up to three photos and finds each dent, scratch, crack, bent, broken, shattered or torn part. Each finding is drawn as a box on the photo, with its part, damage type, severity (minor, moderate or major) and a confidence score.

Compares with the garage estimate

Matches every estimate line against the damage seen in the photos. Lines the photos do not support are listed and flagged for assessor review, because some damage cannot be seen in photos. It also flags damage that the estimate missed.

Calculates a preliminary amount

Adds up the repair lines that match visible damage and deducts the policy excess. A person authorises any settlement. With no estimate, it prices the damage from a built-in rate card for parts, severity, damage type and vehicle model, plus labour.

Recommends the decision

Recommends straight-through approval, confirmed by a person, human review or not eligible, with the reasons. An invalid policy or no own-damage cover is not eligible. Serious structural damage, unseen billed parts, mismatches or missing documents go to human review.

Key Questions Answered

01 / 05

Is it a fraud-detection system?

No. The fraud-cost figure in the film is context only. The assessment checks the policy, the documents and the photos, and sends mismatches to a person, but it does not detect or score fraud.

How it works

  1. 01Upload
  2. 02Extract
  3. 03Validate
  4. 04Assess
  5. 05Calculate
  6. 06Decide

Technology stack

  • Next.js
  • React.js
  • TypeScript
  • Tailwind
  • Gemini API
  • Docker

Use cases

01Motor insurance

Faster own-damage assessments

Assess a claim from the policy, registration document and accident photos, with damage marked on the car and a recommended decision for an assessor.

Want to see AI Motor Claims Assessment on your own claims?

Book a working session

Last updated