Will we need AI to monitor AI?

As we look at technology in this month’s issue, artificial intelligence jumps out as the trend that continues to create the most buzz in the insurance industry. 

One survey, for example, found that 88% of auto insurers currently use, plan to use or plan to explore using AI or machine learning as part of their everyday operations. Seventy percent of home insurers plan to do the same.

AI has been employed across the industry at many levels:

» Underwriting: AI is used to assess risk by analyzing vast amounts of data, including historical claims, demographics, and real-time data such as weather patterns and social trends. 

» Claims processing: AI automates claims processing by quickly evaluating damage, detecting fraud and determining payouts. This reduces the time it takes to settle claims and ensures accuracy.

» Customer service and chatbots: AI-driven chatbots and virtual assistants provide 24/7 customer support. They can answer policyholders’ questions, help with policy changes and even initiate claims reporting. 

» Predictive analytics: AI algorithms analyze data to predict future trends, such as identifying high-risk areas for specific types of claims (e.g., auto accidents or property damage). Insurers can use this information to adjust pricing and underwriting strategies.

» Fraud detection: AI helps detect fraudulent claims by identifying patterns and anomalies in data. It can flag suspicious activities, reducing insurance fraud and saving costs for insurers.

» Risk assessment: Insurers use AI to assess the risk of insuring specific individuals or properties more accurately. This leads to more personalized pricing and coverage options.

» Customer engagement: AI-powered tools personalize marketing and communication efforts, improving customer engagement and retention. Insurers can offer relevant policy recommendations and incentives based on customer behavior.

» Telematics: In auto insurance, telematics devices and AI analyze driving behavior, allowing insurers to offer usage-based insurance policies. Safer drivers may pay lower premiums.

» Data analysis: AI can process and analyze vast amounts of data to uncover insights, helping insurers refine their business strategies, improve operational efficiency and make data-driven decisions.

» Policy management: AI streamlines policy administration by automating tasks like policy issuance, endorsements and renewals, reducing manual errors and administrative costs.

» Risk modeling: AI helps insurers create sophisticated risk models by considering a wide range of variables and scenarios, enhancing their ability to manage risk effectively.

» Health insurance: AI is used to assess health risks, optimize pricing for health insurance policies and support claims processing in this insurance sector.

» Property/casualty insurance: In P/C insurance, AI can assess property values, calculate replacement costs and estimate the risk of natural disasters.

» Reinsurance: AI is being employed in reinsurance to assess and price risks more accurately, helping reinsurance companies manage their portfolios effectively.

» Regulatory compliance: AI can assist insurers in staying compliant with evolving regulations by monitoring and adapting policies and practices.

One of the more interesting uses for AI is developing what is called “synthetic data,” which can be used to improve predicative modeling when, for example, there may be a lack of actual data for certain models. 

Overall, AI has the potential to modernize and make more accurate and efficient many aspects of the insurance industry. At the same time, there are concerns about oversight and regulation — to ensure that its use does not create issues of discrimination, for example. As Senior Editor John Hilton writes about in this issue, the National Association of Insurance Commissioners has been surveying each segment of the insurance industry to quantify exactly who is using AI and how they are using it. The NAIC will use this information in developing AI guidelines for the industry. As rules and regulations evolve, including at the state level, we will see how the oversight of AI develops.

Overseeing AI, however, has a number of wrinkles. There are, for example, “AI hallucinations,” when AI programs make up information out of whole cloth, based on no real data or information. As we see more uses of AI to create images and video, there is more opportunity to move away from reality and create very convincing content that misrepresents the facts.

The next question after regulation is developed, though, is how to monitor AI. Many uses of the technology are so complex, they may require AI tools to monitor and ensure compliance. 

At that point, will we have AI overseeing AI? 

John Forcucci
Editor-in-chief