Which car does the customer drive and how many accidents has he already caused? Was he often sick lately, does he own a property? Does he live in a region where it rains a lot? Insurance companies can find out a lot about their customers, also thanks to social media. You have a huge treasure trove of data. And with driving data from cars, fitness trackers and smart houses, millions of other data sets are added to the companies every day. With this information, insurers could make predictions about which customers will soon suffer damage, who will soon become seriously ill and who might be interested in term life insurance.
This is made possible by new technologies such as artificial intelligence (AI) and data analytics. With the help of algorithms, machines evaluate the large amount of data and calculate probabilities. The insurance industry is one of the industries that will benefit most from these techniques. If companies intelligently evaluate their huge amounts of data, they can assess risks, deal with damage differently and understand their customers better. And they can clear up fraud cases faster.
Customers can also benefit from this. With the help of technology, insurance companies can pay for damage faster because they can plan better. “If a customer submits an invoice for replacement glasses or a hearing aid, no clerk has to look at it anymore, a machine can do it,” says Ina Schneider, head of department at Germany’s largest private health insurer Debeka at a specialist conference of the SZ. “In view of the digitization of all areas of life, customers expect us insurers to respond to their needs quickly and digitally.”
Artificial intelligence is also used in motor insurance: Allianz, for example, wants to quintuple the number of motor insurance claims regulated by AI by the end of the year. The customer photographs the damage, sends the pictures via smartphone, and artificial intelligence calculates the repair costs. However, if irregularities occur when the damage is reported, the systems can sound the alarm. You can see patterns, such as when a customer has previously reported a very similar loss or provided a phone number or email address that played a role in a fraud case. This benefits not only the insurers, but also the customers. Because fewer cases of fraud at an insurer affect the premiums.
However, the algorithms also harbor great dangers: If the machines interpret data incorrectly and draw the wrong conclusions, this can lead to discrimination against certain groups of people.
This is confirmed by Amadeus Magrabi from the start-up Building Minds. The company specializes in AI applications in the real estate industry. “Artificial intelligence is only as good as the data it is fed with,” he says. If training data is included that already reflects social inequalities, or if the algorithm weights certain characteristics such as skin color too heavily, discriminatory decisions can be made.
“Artificial intelligence and data analytics can discriminate against certain customer groups when assessing risk,” agrees Cenk Tabakoğlu, head of the Munich start-up Lumnion, which offers insurers price calculation methods. “Insurers have a lot of data, including from external providers,” he said. “At a certain point, machine learning becomes a black box whose decisions cannot be understood.” Particularly complex algorithms may provide better results, but they carry the risk that the decision cannot be adequately explained. That is why transparent algorithms are necessary.
Petra Hielkema, head of the European insurance regulator Eiopa, also advocated this. Insurers should be held accountable when their AI makes questionable decisions. But they would have to know for themselves how exactly the algorithm works. Hielkema is particularly concerned about possible discrimination against loyal customers. If, for example, a machine recognizes that a customer is very unlikely to cancel, even if the policy becomes more expensive, then the insurer will of course do exactly that: raise the prices. A legal framework is needed, says Hielkema.
This is exactly what is currently being discussed in the EU. The Commission wants to launch the so-called Artificial Intelligence Act, which is intended to better regulate the use of artificial intelligence. The rules should apply equally to all sectors. Hielkema can understand this approach – after all, AI is not only used in the insurance industry. However, she would like the special features of insurers to be taken more into account in the regulation. “The insurance industry is already a highly regulated industry that shouldn’t be over-regulated,” she says. “But we don’t want any gaps either.”