Fraud and Deception

Overview

• In 2023, the discovery of improper auto insurance claims by major used car dealerships with repair shops highlighted the urgent need for insurance companies to enhance their loss investigation systems. Aioi Nissay Dowa Insurance, the lab, and Mind Foundry collaborated to develop an AI model that scores repair shops on fraud suspicion using 4.2 million data points, aiming for transparency and efficiency in fraud detection. This innovative approach, currently focused on revising past agreements and strengthening inspection standards for high-scoring repair shops, has potential applications in various fields to combat fraud and deception.
   

Detail

Background

• In 2023, instances of improper auto insurance claims by major used car dealerships with attached repair shops came to light. As a result, it has become urgently necessary for insurance companies to further strengthen their loss investigation systems

​

Hypothesis

• Aioi Nissay Dowa Insurance possesses a large amount of digital data related to accidents, estimates, and suspected fraudulent claims. AIOI R&D LAB-OXFORD has already established a system for rapid AI development. Mind Foundry has previously developed an insurance claim fraud detection model with Aioi Nissay Dowa Europe and has accumulated expertise in this field.

• Therefore, if these three entities combine their efforts, we can develop a transparent and effective solution in a short period.

​

R&D

• A model has been developed that calculates a fraud suspicion score for each repair shop using approximately 4.2 million past repair estimate and suspected fraudulent claim data, quantitatively demonstrating the basis of these scores.

• The Claims Management Department of Aioi Nissay Dowa Insurance, AIOI R&D LAB-OXFORD, and Mind Foundry are working closely together to design an AI that is trustworthy and effective in its decision-making process.

• The development of this model, including data preparation, model development, application development, and implementation, was carried out in just three months.

​

Next step

• Currently, for repair shops that have received high scores, we are re-examining past agreements and promoting the strengthening of our investigation system, including enhancing the standards for on-site inspections.

• In the future, to adapt not only to changes in vehicle characteristics and repair methods but also to new fraudulent techniques, we will continuously update the AI in this system by re-training it with the latest repair and fraudulent claim data.

• The concept of this solution has the potential to be applied beyond fraudulent auto repair claims, and we are considering its use in combating fraud and deception in various fields.