Video
Status
The solution is deployed within Aioi Nissay Dowa Insurance and is used to support operational fraud investigations.
It received the Claims Initiative of the Year award at the 2024 Insurance Awards in recognition of its innovation in AI-assisted claims management. The model is continuously improved using newly accumulated claims data, investigation outcomes, and repair-cost information, with retraining performed approximately every six months.
FLAGSHIP PROGRAMME
Repair shop fraud scoring
Summary
Using more than four million historical motor insurance claims from Aioi Nissay Dowa Insurance, we developed an AI system that scores the likelihood of fraudulent activity at individual repair shops. The model learns historical fraud patterns and provides both a fraud-risk score and the key factors behind each prediction, ensuring transparency and explainability.
This enables insurers to prioritise investigations while maintaining fast claims settlement for low-risk cases.
The problem
Following a major organised motor insurance fraud case in Japan, insurers needed to analyse historical claims to identify fraud patterns and proactively prevent future abuse.
However, extensive investigations can delay legitimate claims payments, creating a trade-off between fraud prevention and customer experience.
Solution
We developed an AI model that learns historical fraud patterns and produces a fraud-risk score for each repair shop. The system also highlights the factors driving each prediction, improving transparency and supporting investigator decision-making.
By prioritising high-risk repair shops for investigation while allowing low-risk claims to be processed quickly, insurers can improve fraud detection without compromising customer service.