Automation as a driver for digitization and change

Data processing lies at the heart of the insurance business. Insurers collect and process personal data for several reasons. This can range from analyzing risks that customers wish to cover, paying claims and benefits, through to detecting and preventing fraud. For insurance companies, regardless of their stage of digitization, automated decision making has been, and largely remains, a hotly debated area. This is not just because new entrants keep disrupting some of the established views amongst domestic firms, but mainly because automation has the capability of cutting costs and increasing shareholder value. 

Overview on digital sales

Digitization, i.e. the process of changing from analog to digital form, as well as digitalization, the use of digital technologies to change a business model and provide new revenue or value producing opportunities, has arrived in the insurance industry. It is not only consumers that benefit from the digital transformation of conventional insurance services but, thanks to the rise of insurtechs, completely new business areas and insurance products are created. In keeping with what has been happening in other areas of financial markets, insurance policies are increasingly sold online. In 2017, total gross direct written premiums in Europe increased 4.7% to €1 213bn¹, with relatively strong growth in all business lines². This robust growth was mainly driven by the three largest European markets (UK, France and Germany).

Even in Germany’s relatively “traditional” consumer market, digital disruptors and new entrants have moved the “established” insurance industry to embrace insurtechs, both as part of new business distribution channels but equally as a way of engaging more generally with their clients. Direct sales in Germany accounted for 15%³ of new business in property and casualty insurance and 7.3%⁴ in 2018. Direct health and life insurance sales accounted for 2.2% of new business, showing no increase⁵. German-based consumers increasingly switch to simpler, more user-friendly products, even if that means changing providers. Part of that optimization has meant moving to automated decision-making, artificial intelligence and machine learning as part of the value and client engagement chain.

From disruption to collaboration

This optimization certainly makes for an interesting time of change in the domestic market, where competition steps up amongst existing and new entrants, as they bid to attract and retain consumers spoiled for more choice. The insurance industry as a whole is seeking to become more agile in how it helps consumers with purchasing or switching products. Another competitive advantage is automated claims-handling processes, which a newly founded and fully licensed EU insurer has recently introduced to the German market on a cross-border basis⁶. At the same time, a couple of incumbent insurers are tackling the harder part of updating legacy back-end systems, whilst at the same time they have to catch up with the recent trends in digitization and customer centricity. Their need to keep track with market developments paves the way for collaboration with insurtechs. 

General GDPR requirements

Division of responsibilities under the GDPR

To ensure compliant contribution to insurance value, including by using automated decisions in a highly regulated industry, all providers must ensure their data is properly collected and evaluated. Collectors and data controllers must act in line with the EU’s General Data Protection Regulation (GDPR)⁷. When insurers collaborate with insurtechs, the role of the controller as per Art. 4 No. 7 GDPR does not have to be assigned to the insurer from the outset. Rather, this will depend on the concrete design of how the parties document and operationalize their collaboration.

Depending on the concrete design of the collaboration, insurers may consider outsourcing an insurance activity or important function. Subject to the terms of such an agreement, this may shift the role of the data controller to the service provider. In other cases where insurers give on-going instructions to their service provider (processor) and determine the means and purposes of data processing, they remain data controllers and need to comply with several obligations. The GDPR not only obliges them to implement appropriate security measures and data protection policies, but also to actively demonstrate that they are compliant with all regulatory provisions. Individuals and supervisory authorities can hold both controllers and processors liable if they fail to comply with their responsibilities.

Furthermore, the GDPR has introduced new rules on joint controllers. Joint controllers together decide on the purposes and means of data processing, i.e. they pursue an identical or a shared purpose. Controllers will not be joint controllers if they process the same data for different purposes. According to ECJ case law⁸, joint controllers may be involved in the processing of personal data at different stages and to different degrees in such a way that their degree of liability will depend on all relevant circumstances of the particular case. Again, it will depend on the concrete design of a cooperation whether the GDPR rules on joint controllers apply.

Automated decisions on individuals, legal effects and significant impairment

If a data controller, be it an incumbent insurer or an insurtech, uses a machine in a decision making process, the issue arises whether such a decision, including its underlying data flows, actually qualifies as automated individual decision making (AIDM) according to Art. 22 (1) GDPR. The GDPR stipulates that AIDM is only permissible for the purposes of concluding or administering a contract (Art. 22 (2) lit. a) GDPR) and may require the customer’s explicit prior consent in some cases (Art. 22 (4) GDPR).

Automated decision making often involves profiling, but it does not have to. Profiling analyses aspects of an individual’s personality, behavior, interests and habits to make predictions or decisions. Although many people think of marketing as being the most common reason for profiling, this is not the only application. Profiling and automated decision making can be very useful for organizations and benefit individuals in many sectors, including healthcare, education, financial services and marketing. They can lead to quicker and more consistent decisions, particularly in cases where companies must analyze a very large volume of data and decisions very quickly.

If a system evaluates personal data without the involvement of a person, the question becomes when such an evaluation has legal effect. Such legal effect requires that a person’s legal status or legal rights actually change. For example, there is no legal effect when only undoubted legal powers are exercised, e.g. when conducting access control on employees by means of chip cards when they enter a company’s premises. The same applies to the service of documents when instituting court proceedings or judicial default action. Although these decisions are implemented automatically, rights are not changed, nor fundamental rights infringed without justification. Hence, not every automated process is an AIDM from the outset and thereby subject to specific GDPR requirements.Rather, automated processes must achieve the following in order to qualify as AIDM. They must:

• evaluate someone’s personality traits beyond a mere “if-then decision”,
• modify an existing legal position,
• have a minimum impairment on the freedom of the person concerned. 

Automated decision making in the digital insurance space 

Property and liability insurance

In the property and casualty space, data processing only comprises family, master and contract data. The customer submits his application online; the insure will reach his decision (rejection or acceptance) through an underwriting tool, without the involvement of an underwriter. The customer may also submit his claim online. The insurance company will decide on the claims settlement based on a specific algorithm. In practice, this algorithm will often make it possible to settle claims below certain thresholds without further verification. When underwriting risks or deciding on claims, the insurer assesses the personal circumstances of the applicants and thereby changes his legal position according to Art. 22(1) GDPR. Automated underwriting and claims decisions qualify as AIDM and require justifications under the GDPR.

This justification derives from Art. 22 (2) lit a) as the AIDM is necessary for entering into (underwriting) or performing the insurance contract (claims handling). However, the insurer or the insurtech, depending on who is the data controller, must implement suitable measures to safeguard the applicant’s rights, freedoms and legitimate interests. The least that the data controller must provide for is the right to obtain human intervention, to express an opposing point of view and to contest the AIDM. In practice, this will require an efficient on-line process to have the AIDM reviewed by a senior underwriter or claims manager in case the insurer rejects cover or the payment of a claim. Thus, in the area of property and liability insurance, the GDPR protects the rights of customers by granting a right to control the AIDM in hindsight.

Life and health insurances

In the life and health insurance space, the underwriting of risks and the management of claims necessarily involves the processing and administration of health data (Art. 4 No. 15 GDPR), not just contract data. The customer journey usually takes place via a sophisticated website through which the customer submits his application. Should the underwriting decision not require the involvement of an underwriter, an underwriting tool will make the decision. The underwriting tools currently used on the market base their decisions on a process, which initially queries standard constellations such as simple health questions and questions on leisure habits and then transfers the applicant's answers to a multi-layered decision tree. Within each layer, the algorithm either reaches a decision or transfers the decision to an underwriter. The customer may also submit his claim online. Depending on the complexity of the claims, the insurance company may then decide on the settlement of the claim through specific algorithms.

However, when underwriting risks through tools or deciding on claims based on a certain algorithm, the insurer assesses a customer’s personal circumstances and thereby changes his legal position pursuant to Art. 22(1) GDPR. Thereby, automated underwriting and claims handling decisions also qualify as AIDM and require specific justifications for life and health insurance. 

Contrary to non-life policies, AIDM requires the customer’s express and prior consent. Insurers and insurtechs should seek required consents through sophisticated websites to document those through their login files. This documentation will safeguard their position in case of future disputes with customers or regulators. Just like non-life insurance, the data controller must provide for an appropriate process ensuring that a senior underwriter or claims manager checks the AIDM.

Distinction between contract data and health data

Because of the different AIDM requirements for life and non-life insurance, it is worth looking at the distinction between contract and health data.

According to the GDPR, health data is personal data resulting from specific technical processing relating to the physical, physiological or behavioral characteristics of a natural person, which allows or confirms the unique identification of that natural person, such as facial images or dactyloscopic (fingerprint) data (see Art. 4 No. 15 GDPR). This also includes further personal information about a customer, e.g. a number, symbol or particular assigned to uniquely identify him for health purposes⁹. Fitness apps are an important use case for this definition. The Art. 29 Working Party Group, i.e. the predecessor of today’s European Data Protection Board, was of the view that a distinction should be made between raw data (e.g. number of steps on a treadmill) and analytical data (e.g. the conclusions to be drawn). This position should still be valid, provided the data controller does not use the data to incentivize benefits of the insurance product and the data is sufficiently anonymous to ensure the data controller can only draw conclusions about the user of the app with extreme effort. This effort should be measured by the costs spent on the identification, the time required for such identification and the technology and technological developments available at the time of processing. In contrast, the purely hypothetical possibility of carrying out an analysis of health data should not be sufficient. 

Outlook 

In summary, the GDPR, welcomingly, permits AIDM provided it is designed and conducted in a manner that properly reflects the legal requirements. Used correctly, automated decision making is useful for many businesses. In the insurance space, it can help to interpret policies correctly and make decisions fairly and consistently. The GDPR recognizes this and does not prevent companies from carrying out profiling or using automated systems to make decisions about individuals from the outset, but provides for rules ensuring appropriate protection of private individuals. 

For future market players this may mean assessing existing policies and processes, but also developing new documented ones that support and explain the “tech” in AIDM, both to regulators and consumers, while at the same time ensuring such changes do not limit the user-friendliness and the advantages provided by digital distribution. The unstoppable digitalization opens up new opportunities and possibilities for the insurance industry − but platform providers should not underestimate the requirements for legitimate automated decisions and the data processing associated with it.

If you would like to discuss strategic options or any of the items mentioned above, in particular how to plan ahead for any impact on operational compliance in both documentation and non-documentation work streams or how these priorities may affect your business or your clients more generally, please contact our Eurozone Hub key contacts.

1. See Insurance Europe’s Report on European Insurance in Figures 2017 Data.↩
2. Life +5.0%, health +4% and P&C +4.4%.↩
3. Increase from 13.9% to 15.0% - German Insurance Association statistical handbook of the insurance industry 2018 p. 13.↩
4. Increase from 6.7% to 7.3%, supra↩
5. Stays constant between 2.2% and 2.3%, supra↩
6. https://www.lemonade.com/de/schadensabwicklung↩
7. See Regulation (EU) 2016/679 of the European Parliament and of the Council of 27 April 2016 on the protection of natural persons with regard to the processing of personal data and on the free movement of such data, and repealing Directive 95/46/EC (General Data Protection Regulation “GDPR”), in force since 28.05.2018 here.↩
8. EuGH C-210/16 dated 5.08.2018 („Facebook“); EuGH C-25/17 dated 10.07.2018 („Zeugen Jehovas“); case decided before GDPR.↩
9. Recital 35 GDPR.↩