Every industry is exploring the uses of artificial intelligence and machine learning and the automotive insurance sector is no exception.
Doug McElhaney, a partner within the insurance practice at McKinsey & Company, explains that it’s not just about AI algorithms but also the data that’s being incorporated into them: “For the U.S. market most carriers will use linear algorithms and they are using a body of information – 20 or 40 pieces of information to separate high versus low risk. Advanced, non-linear algorithms can identify at a more granular level that may be missed by that rudimentary algorithm. All of AI is designed to replicate human cognition to some degree. More advanced algorithms associated with machine and deep learning do this more so than linear algorithms such as linear regression.”
He adds that with advanced AI algorithms are better at discerning the different levels of risk. An example would be the tiering of risk that’s done in automotive insurance. With an increased amount of data available to assess risk, together with advanced algorithms, there might just be an insurance classification with only 10 risk classes.
Telematics-based Insurance
Telematics-based automotive insurance products offer access to increased volumes and types of data. The data capture emanates from the vehicle itself, or from a mobile that’s being used within the vehicle. Data can be captured at high velocity and the data stream can identify specific driving behaviors, such as rapid turns or hard braking.
McElhaney elaborates: “The presence of telematics products offers carriers increased confidence. With a telematics-based product, the carriers have moved from a static rating approach to a usage-based insurance product. I capture information from you using a device that gives more information about how you drive. It’s a different type of information as compared to more traditional static rating factors.”
Challenges and Risks
Roman Swoszowski, vice-president, AI and cloud research and development at Grape Up says: “To capture, use and analyze data poses significant challenges and risks for insurers because automakers may leverage their data competence to enter or disrupt the insurance market.” This may mean offering their own insurance products or by partnering with selected insurers, thus reducing the choice and bargaining power of customers and intermediaries.
He also claims there is a risk of demand being reduced. The profitability of insurance products might be, too. Yet data -driven technologies potentially could improve vehicle safety, reliability and efficiency. This may also lead to fewer and less severe accidents, lower repair costs and subsequently lower premiums.
He adds: “On the other hand, it offers insurers access to more accurate and granular data on vehicle performance, driver behavior and accident scenarios, which can improve risk assessment and pricing as well as enhance customer segmentation and personalization.”
Expanding Data Sources
Matthew Carrier, principal at Deloitte Consulting, says a multitude of expanding data sources are being used by insurance companies. Telematics data is but one of them. Other data points include traffic patterns and accident information by location, such as intersections, roads and highways. Having this information at hand can enhance the ability of insurers to assess risk and the pricing of insurance policies.
He adds: “Information gleaned after an accident occurs, such as vehicle camera video and telematics data at the time of an accident, can be used for processing a claim. Additionally, partnerships between insurance companies and automakers can streamline auto repairs by identifying required parts and availability.”
Driver Behavior
Artificial intelligence and machine learning capabilities improve the detection of fraudulent insurance claims or purchases. This may include factors such as the expected miles that will be driven annually, the garage location of where a vehicle is parked, or the extent of someone’s injuries after an accident or appropriate medical treatments. “The efforts by insurance companies to reduce fraud ultimately benefits consumers by lowering the cost of insurance policies,” explains Carrier.
Ethical Considerations
As for ethical considerations, he says there are many. They are focused on confidentiality and privacy, bias, and infringement as it relates to consumer data. There is a need to ensure that the consumer data that’s available for any risk assessment is accurate, appropriate and consistently collated. There is also a need to observe consumer protections, laws and regulations, and there needs to be some consideration as to whether the way the data is being used leads to biases.
