Fraud affects every industry, and insurance is no exception. False declarations to the subscription, false claims, and amplification of the disaster cause insurers to lose several hundreds of millions of dollars every year because of frauds. The FBI reports that the total cost of insurance fraud is estimated to be more than $40 billion per year, excluding medical insurance. The extra expense for insurance companies costs the average U.S. family between $400 and $700 per year in the form of increased premiums.

Artificial intelligence (AI) is a key watchdog for insurance fraud detection and prevention and fights against fraudulent claims. AI applications for insurance have only been made possible in the last few years with the explosion of massive client data access and computational resources. Through this article, we describe some of the many ways in which AI can cut down on insurance fraud.

Detect fraudulent claims:

AI algorithms analyze the data on insurance policies and claims. Shift Technology, a French start-up working towards automating claims, can identify as many as 75% of fraudulent claims by aggregating and cross-checking the data of several undisclosed partners, such as insurers, which helps insurers speed up processes for bona fide customers. The Shift Technology AI model compares multiple seemingly unrelated claims to detect unusual similarities in circumstances or invoices and detect statistically unlikely trends of claims to raise the red flag for human investigation. The AI algorithm evaluates the degree of doubt on a file and associates a score. Beyond a certain threshold, the fraud service of the company takes over.

Car accident fraud detection is another field that AI is revolutionizing. By automating the claims process, American start-up Galaxy.AI, can estimate the damage cost to automate the property and casualty insurance claims process. This enables their clients to submit image-based claims using their mobile phone and get an immediate value-added response (as seen in the image below). Therefore, the AI algorithm keeps learning the severity of an accident and can later determine if accident images and the cost claimed by the client correspond.

Predict and control risks:

Many drivers report car accidents that are unrelated to their own distraction and carelessness on the road. This type of behavior is a fraud, and AI technology can help to decide if this is the case. This is a particularly important optimization proposed to insurers. Safety driver monitoring systems use AI models to detect the driver’s attention in real time. Monitoring the driver’s attention level with visual recognition systems would help determine the level of responsibility and predict risks more quickly, especially in the event of fatigue or elevated blood alcohol levels.

This video features a vehicle using the Renovo AWare platform and driver monitoring software from its AI partner Affectiva to detect driver distraction. A series of thresholds are set for escalating levels of driver assistance.

It is necessary to establish a policy framework regarding the ownership, use, and protection of data collected by autonomous vehicles, including safety driver monitoring systems. Such a framework is still under development and could include various issues such as:

• policy development regarding the need for governments and other stakeholders to use this data, while protecting the privacy and security of user data;
• access to maps generated by autonomous vehicle companies for use by local governments such as urban planning, engineering, and traffic management;
• ownership and control of data on autonomous vehicle users and the possible role of integrated privacy protection and other frameworks; and
• the transparency of automotive technology.

Additionally, AI can characterize driving behavior using automatic visual analysis. In fact, rash drivers generally tend to accelerate quickly and in forward direction. They change lanes frequently and get dangerously close to other vehicles and people.

Using real-time video streaming, the AI model takes two consecutive frames, extracts features that will characterize the driving behavior and threshold them, and determine if rash driving is detected. Furthermore, the alliance of AI and visual technologies such as drones could revolutionize the treatment of claims by recognizing the type and severity of a road accident present and compare with those already observed and whose data were recorded. In addition, this technology will eliminate litigation between insurers and insured.

Predicting and controlling the risks associated with driving connected vehicles would thus significantly reduce compensation costs and thereby improve the insurer’s financial results. Self-driving prediction algorithms are still being developed in an academic setting.

Leverage web metadata:

Conclusion:

According to a study published by McKinsey Global Institute in 2017, nearly 43% of transactions in the insurance sector could be completed using AI. While the arguments for harnessing the power of data are convincing, there are still many functions that AI cannot perform due to its limitations, which raises regulatory concerns and customer acceptance, inhibiting the adoption rate. The establishment of an insurance system based on AI technologies is not without risks.  In the event of an error in the AI fraud estimation, the responsibility of both the insurer and the insured is still an open question.

If you have any questions or would like to know if we can help you with your innovation challenge, please contact our Strategic Accounts Manager, Linda Cohen at lcohen@prescouter.com.