[Skip to content]

Sign up for our daily newsletter
The Actuary The magazine of the Institute & Faculty of Actuaries

Buried treasure: Could data regulations sink AI in Europe?

Strict data regulations may hinder the role of AI and data analytics in bringing innovative insurance products to the European market, warns Frank Schiller



“The use of big data and AI has not only provided a new, more attractive customer interface – it has also completely revolutionised products themselves”

The wealth of data held by insurers is a treasure trove. Technologies such as AI and data analytics are enabling the industry to create a world of products that can adapt more individually and more quickly to customers’ needs than ever before. Strict data protection requirements still give life assurers in Europe a certain market advantage – but for how long?

Daniel Schreiber, CEO of Lemonade, had a provocative message for the insurance sector in 2018: “The next insurance leaders will use bots not brokers, and AI not actuaries.” Some other sectors are already on the home straight in this journey. Airbnb makes exclusive use of a digital interface to arrange customers’ accommodation, and Uber has completely revolutionised taxi booking services via comprehensive digitalisation, automation and the involvement of a wide community. The use of big data and AI has not only provided a new, more attractive customer interface – it has also completely revolutionised products themselves. What would the consequence be if we were to apply these concepts to life insurance? In view of the strict regulation of the insurance sector in Europe, is this even possible?

China leads the way

With Ant Financial as the product provider, Chinese conglomerate Alibaba launched a revolutionary product in autumn 2018, covering specified services for around 100 critical illnesses. When the contract is concluded, a simplified risk assessment, based on Alibaba’s Sesame Credit point system and a general age differentiation, is carried out; there is no medical risk assessment. Sales and claim assessments are handled online and, if required, other Alibaba customers are involved to help identify fraud. The premium is not guaranteed in advance; instead, the actual claim payments incurred across the board are charged to all customers in equal fortnightly instalments.

The product itself has been kept very lean. Policy issue and the operation of the policy are linked to existing processes at Alibaba and are digital. Claim assessments are carried out digitally by a small team, with the help of the community where required. Everything is done digitally, implementing large elements of Schreiber’s vision. 

After just nine days, 10m customers had bought the product. After six months, the figure was more than 65m, around 10% of Alibaba’s customer base. Of these customers, 62.5% had never before considered buying a comparable product.

The product’s formula for success is its combination of simplicity and transparency. Every customer with at least 650 Sesame Credit points can buy it. Ongoing contributions depend solely on claims paid out. Even these are transparent, and each individual has a certain amount of influence.

The second pillar of success is Alibaba’s excellent market penetration. This is the only reason the company was able to launch this experiment and adapt the product over time to make it even more successful.

Ultimately, the large number of customers and high evaluation frequency (24 times per year) enable the product to be improved continuously based on sound data. As a result, it is optimally aligned with short-term customer requirements and the economic situation.

Transferring the concept to Europe

For companies in Europe, there are several requirements and problems to be addressed when transferring these solutions. 

Supervisory requirements

The European Insurance and Occupations Pensions Authority (EIOPA) has clearly expressed its expectations regarding products based on big data. Earlier in 2019, EIOPA chairman Gabriel Bernandino said: 

“Insurers have to adapt their governance frameworks to address the challenges posed by these new technologies, in particular issues with the fairness of the use of big data analytics and the accuracy and explainability of ‘black-box’ algorithms.”

In terms of processes, the supervisory authorities are clear that responsibility for the results of processes based on big data and AI cannot be assigned to machines. When insurers are designing (partially) automated processes, it is important that they guarantee these processes are grounded in an effective, appropriate and proper business organisation.

In addition, the aim of supervisory authorities is to keep financial services access open to as many people as possible. Consumers of AI and data analytics applications present a number of regulatory issues for insurers. If there is an entry threshold (such as the Sesame Credit points), it must involve no implicit illegal discrimination based on, for example, gender or origin; therefore, a threshold value cannot be derived purely using algorithms based on the available data. It is necessary for the results to be verified and corrected by experts – in insurance, typically actuaries.


Click here to enlarge 

Special protection for health-related data

Compared to China, there are significant restrictions on data processing in the life and health insurance sector in Europe. According to the EU’s General Data Protection Regulation, processing of personal and health-related data requires the highest level of protection. Within a company, personal and medical data can only be evaluated on a need-to-know basis and when the customer has consented to processing of the data. It is prohibited to pass the data to third parties.

Data volume and evaluation frequency

Some features of life insurance make it unsuitable for the use of big data, at least in the short term. In terms of the key dimensions of big data, these are: 

Volume: The data volume should be huge – but the number of death or critical illness claims tends to be just a few hundred per year

Velocity: It should be possible to generate and process the data quickly – but in insurance the frequency tends to be annual

Variety: The data should be very heterogeneous to enable robust predictions to be made – but insurance frequently has very narrowly defined target groups

Veracity: The data quality should be good – but insurers frequently have to battle with legacy systems for their data

Value: Ultimately, the product has to be an economic success.

Actuaries will play a major role in a data-driven business model for insurers. They will be needed in particular for the pooling, extrapolation and checking of data, assumptions and models, as well as interpretation of the results (see Figure 1).

Missed chances

For life insurance in Europe, regulation is both a support and an additional burden. It is difficult for providers such as Amazon – the Western equivalent of Alibaba – to develop and market insurance solutions directly through established sales channels and using their superior technical methods in the European regulatory environment. The regulatory requirements in data protection and the additional business organisation necessary make it unattractive for them to become directly active in the European life insurance market.

As a result, some life assurers feel protected, invest little in their own digital developments and show minimal interest in experiments. They are missing out on the chance to become purely digital players and transform into agile companies that quickly adapt to customer needs and the economic environment.

Major tech companies such as Uber and Airbnb have set the example in their fields, but life insurance players will soon come up with a comparable market dynamic. Only agile insurers will then have sufficient resources and data – and the required expertise – to play an active role in this environment.

Dr Frank Schiller is chief actuary for life and health reinsurance, Europe, Latin America and Middle East at Munich Re