Using Artificial Intelligence in Non-Life Reserving

The use of artificial intelligence (AI) has skyrocketed in many areas. The insurance industry, for example, is a domain where technological innovation and traditional practices come together. In this blog, we will discuss the current status of AI use in insurance, go into the innovations in Non-Life reserving that are within reach of practical application, and set out the challenges that need to be addressed when applying AI. In future blogs, we will discuss other applications of AI, such as pricing in insurance.

In the insurance industry, artificial intelligence has emerged as a transformative force, reshaping traditional practices and enhancing operational efficiency. The utilization of AI in underwriting has evolved, with predictive analytics and machine learning algorithms analyzing extensive datasets in real time to create more accurate risk profiles. This not only allows insurers to tailor their policies to individual needs but also ensures fairer and more competitive pricing. Claims processing, often perceived as a cumbersome task, has undergone a revolution with AI-driven applications. Natural Language Processing (NLP) algorithms and image recognition technology expedite claims settlement, providing faster and more accurate assessments. Customer service has been revolutionized through AI-powered chatbots and virtual assistants, offering instant responses and improving overall responsiveness. Additionally, AI's role in fraud detection has become indispensable, with machine learning algorithms identifying patterns indicative of fraudulent activities, thus contributing to a proactive approach in maintaining the integrity of the insurance ecosystem. These AI applications have not only improved accuracy, efficiency, and personalized customer experiences but also set the stage for ongoing innovation in the insurance landscape. The current state of AI in insurance signifies a pivotal moment, laying the foundation for a future where technology continues to redefine industry norms and elevate customer satisfaction.

The application of machine learning or AI in general for Non-Life reserving has been limited so far. The most important limitation has been data availability. The collection of data has been improved over the past years; more structured data and more unstructured data are collected. Also, computer power has increased, and science has brought new insights into machine learning and deep learning. 

As computing power has increased and more data has become available, possibilities arise when applying AI to Non-Life reserving. Machine learning and more specific deep learning approaches can be applied. This applies to both case-by-case reserving and actuarial estimations.  

Currently, updating case reserves requires significant manual effort. Furthermore, parts of the process are outsourced, which makes frequent updates cumbersome. Looking specifically at bodily injury claims, case files contain expertise reports and unstructured data. This information is used for obtaining case reserves estimates. The actuary is using data at aggregate homogeneous risk group level estimating future payments, including payments for claims that have occurred but have not yet been reported. For claims with a long settlement period, it is important to use as much information as possible for accurate estimates. Therefore, it is attractive to include information from individual files.  However, when a group of claim handlers changes its reserving approach, incurred claim patterns are disturbed. Automating the entire process including case reserves will remove this issue. Machine learning offers the possibility of including public data sources, e.g., for the estimation for future career, which is used for the calculation of income loss that is compensated in a bodily injury claim.

The process of introducing AI for claims reserving is not trivial. The first areas where AI could be applied in practice are simple material damage claims and possibly short-duration bodily injury claims. Losses resulting from more complicated bodily injury claims depend on many parameters. Information is available in case files partially as structured and partially as unstructured data. The technology interpreting the relevant unstructured data is not mature yet. It can be tested and further developed in other areas than claims reserving. For example, AI can be used for interpreting policy conditions. This can first be applied internally by insurers, for example, by employees in a call center. When technology is proven, it can be used by chatbots communicating with clients. Insights gained from these developments can be used to improve the claims reserving process using unstructured data.

Using AI has the potential of making the reserving process more accurate and more efficient. In the context of insurance companies becoming more efficient using new technologies, this is an essential step in staying ahead of the competition.

In our market observations, we identified several challenges to using AI within the insurance sector. Here are two important ones:

• Data Quality and Availability. Investment in data collection is critical. Because historical data is required, additional data is not directly paying off. The future effectiveness of AI models is heavily dependent on the caliber of data they process, with poor data leading to unreliable AI outcomes.
• Ethical Implications and Bias. AI systems can unintentionally perpetuate biases, potentially leading to unfair treatment of certain customer groups. In reserving, this is less critical compared to pricing. Therefore, reserving is an attractive topic to start using AI.

These challenges highlight the need for a careful and thoughtful approach to integrating AI in the insurance sector.

As we conclude our exploration into the role of AI in the insurance industry and more specific in Non-Life reserving, it's clear that we are witnessing a pivotal moment in its evolution. There are many challenges, such as data quality and ethical concerns, but the potential for innovation is immense. Reserving for Non-Life is an example where improvement is within practical reach. We invite you to engage with these advancements, explore AI's capabilities in your practices, and participate in discussions shaping a more efficient and equitable insurance landscape. Staying ahead of the curve is both fun and an economic necessity.