De Nederlandsche Bank is closely monitoring developments regarding the application of Artificial Intelligence in the insurance sector. Previous research - 'AI in the insurance sector' (2018) and 'DNB Theme Research Insurtech' (2019-2020) - already showed that the use of advanced data analytics and AI continues to increase. The regulator therefore thought it was high time to re-inform itself about the current state of affairs. DNB therefore spoke about the application of AI with six medium and large insurers and provides a brief review.
From the interviews conducted, it appears that insurers are making increasing use of (advanced) data analytics and AI. According to DNB, the insurance sector is one of the industries that has traditionally made extensive use of data, particularly in the actuarial domain for the purpose of setting premiums and estimating adequate provisions.
The regulator emphasizes that - given the relatively small selection of insurers - the results do not necessarily provide an overall picture of the sector, so the conclusions drawn should be interpreted with caution. And where DNB uses the terms "AI" and "advanced data analysis" to describe developments, the regulator does emphasize that these terms are difficult to delineate.
Those surveyed indicated that the use of data analytics is increasing in all parts of the value chain (including underwriting and claims handling). "At property-casualty and health insurers, this is visible throughout the value chain. At life insurers, the focus is particularly on automating and optimizing operational processes, for cost efficiency. The application of advanced data analytics is seen as especially useful if it is scalable and repeatable."
Detection of anomalous claims (fraud detection), email routing to appropriate mailboxes, "robo advice" and chatbots, and analyzing customer texts using text mining were mentioned in the interviews as examples of using AI.
In addition, insurers indicate that the frequency of existing analyses (e.g., modeling risk premium) has increased significantly in recent years. Internet of Things applications are not yet widely and widely deployed, although some applications and pilots are underway. Insurers see the potential of this particularly in prevention.
No examples emerged in the interviews of fully automated decisions based on AI algorithms, without any human intervention. According to DNB, insurers seem reluctant to deploy such "black-box" models. The applications of advanced data analytics and AI are mainly used to support and advise in making decisions.
The surveyed insurers indicate that human intervention in the process is important. For example, in the case where an algorithm detects an anomalous claim (possible fraud). In such cases, the final assessment is always done by a human.
Insurers expect advanced data analytics to become more and more common. They also expect the techniques to make it possible to increasingly refine premium setting, however, this does raise the possibility that advanced risk selection could lead to uninsurability. Accordingly, in the interviews, insurers expressed concerns about such risks associated with further premium differentiation.
Finally, DNB states that it continues to closely monitor technological developments in the insurance sector. To this end, the regulator has three follow-up steps planned for 2022. These include continuing the project "AI and explainability," seeking better coordination in terms of supervisory activities with the AFM to avoid overlap, and continuing DNB's involvement in international consultations on AI (e.g., through EOIPA).
