What challenges did Autonet face before AI implementation?

Manual car insurance damage assessments were slow and inefficient, delaying claim processing and customer satisfaction. Existing pre-trained image recognition models could not accurately classify vehicle damage types and severity needed for claims processing.

What was the impact of the AI solution?

The solution enabled fast, accurate damage classification, reducing claim processing times significantly. Customers often get their car repaired within two days, improving customer experience. Sending cases to SMART repair specialists saves at least $500 per claim, reducing costs and environmental impact. The integrated workflow provides real-time case visibility and collaboration with repair partners.

Source

Evidence: Low25/100

Autonet Automates Car Insurance Claims Processing using Amazon Rekognition Custom Labels

Use case typeFault detectionUpdated Jun 13, 2026

Autonet, part of Progrits AB, operates in the Auto Insurance industry and aimed to automate the slow and manual car insurance claims damage assessment process. They used Amazon Rekognition Custom Labels to build a highly accurate model for damage level classification with significantly fewer annotated images. The model was integrated into a workflow allowing customers to upload photos of their car damage to get quick and accurate assessments. The assessment results help direct claims either to SMART repair specialists for minor damage or to traditional body shops for more severe damage. The partner repair shops are integrated into the workflow for case handling and final approval of repair category.

Organization: Autonet
Industry: Insurance
Location: Sweden
Published: June 2021

Reported outcomes

Strategic outcomes

Speed & agilityAutomated and accelerated claims assessmentsCustomer experience & trustImproved repair turnaround for customersCost efficiencyReduced claim handling costsNew product / capabilityEnabled detailed damage classification

Why do we believe this deployment?Customer identity, provider attribution, maturity, and source checks

Customer: Autonet
Provider: AWS
Maturity: Unknown
Linked source: AWS Machine Learning Blog

No explicit deployment-stage evidence found.

Customer identity supportedSource describes one deploymentMaturity evidence evaluated

Primary read

Use case focus

Showing 1 of 1

1Damage Detection

Manual car insurance damage assessments were slow and inefficient, delaying claim processing and customer satisfaction.
Existing pre-trained image recognition models could not accurately classify vehicle damage types and severity needed for claims processing.

Autonet leveraged Amazon Rekognition Custom Labels to train a custom damage detection and classification model using about 10,000 annotated images.
The model utilized transfer learning from a base model trained on millions of images for higher accuracy with less data.
Customers upload photos of car damage which are analyzed first by Amazon Rekognition pre-trained labels API for general tagging, then by the custom damage model for detailed damage classification.
The results are combined with vehicle configuration data in a data model to predict appropriate damage repair categories.
The workflow routes cases to SMART repair partners or full body shops depending on damage severity, with partner final approval.
This integrated AI-driven workflow automates and accelerates car insurance claims assessments and repair routing.

Technologies

Amazon Rekognition Custom Labels Amazon Rekognition pre-trained labels API

The solution enabled fast, accurate damage classification, reducing claim processing times significantly.
Customers often get their car repaired within two days, improving customer experience.
Sending cases to SMART repair specialists saves at least $500 per claim, reducing costs and environmental impact.
The integrated workflow provides real-time case visibility and collaboration with repair partners.

Architecture

Images are captured from the user's device and first sent to Amazon Rekognition pre-trained labels API for general tagging. Then images are sent to a custom damage detection and classification model trained with Amazon Rekognition Custom Labels that assesses the damage location, type, and severity. Damage detection results are shown to customers for confirmation. The combined damage data and vehicle configuration information is input into a downstream data model to predict repair class. Cases are routed to SMART repair partners or traditional body shops accordingly, with partner approval integrated in the workflow.

Sources & evidence1

Evidence: Low25/100Evidence strength