Artificial intelligence in insurance: Targeted marketing as a quasi-underwriting function

AI is causing dramatic shifts in the insurance industry. Artificial intelligence (“AI”) can analyse large amounts of data and make very accurate predictions, which is naturally already leading to insurers increasingly using AI to identify risk and drive more accurate pricing.

To further enhance these advantages of AI, some businesses are also starting to use targeted marketing in ever more sophisticated ways in an attempt to ensure their products are marketed more effectively. Whilst these new techniques can provide real benefits to insurers, they do come with additional risk, particularly in relation to protected characteristics and discrimination. 

What is Targeted Marketing?

Targeted marketing is a strategy used by businesses to increase customer sales through the use of highly personalised advertising. It involves analysing large amounts of customer data and marketing trends to identify specific audiences. This approach allows businesses to generate tailored marketing content specifically designed to resonate with the individual. The use of AI in this process is increasing the accuracy with which specific types of customers can be targeted and the effectiveness of the adverts they are presented with. 

The line between targeted marketing and underwriting 

Customer data used in targeted marketing may include protected characteristics such as age, race and gender or algorithmic proxies to protected characteristics. Algorithmic proxies are factors which are predictive of or correlated with other characteristics, which may be protected. 

In England, insurers cannot legally use a number of protected characteristics in their underwriting of risks. For example, since 21 December 2012, insurers have not been able to use gender as a rating factor in pricing. It is important to note that this ban does not mean that insurance prices are gender neutral, as other factors influence insurance pricing such as algorithmic proxies to gender. This means that some gender related risk is still factored into some insurance prices. There are significant differences between the accident risk, morbidity risk and mortality risk between men and women. This means that the cost of providing insurance products that cover these risks to men and women are different. For some insurance products, gender was the second most significant risk factor after age, before the 2012 ban, for example, for term life insurance products – this is where a lump sum of money is paid out to named beneficiaries if the insured dies. Mortality statistics show that gender is a key factor in the probability of a person dying; the mortality rate increases with age and for all ages it is higher for men than for women of the same age. In a legal environment where gender cannot be used as a rating factor in the underwriting process, targeting marketing towards the lower risk gender can result in clear cost saving benefits and higher profits, especially in cases where gender is highly correlated with risk.

Insurers use targeted marketing and distribution processes to control the gender mix in their insurance portfolios. As insurers use increasingly powerful and effective AI tools to target the more profitable gender in marketing, targeted marketing may increasingly become a quasi-underwriting function. The legal implications of using protected characteristics through the backdoor of targeted advertising are uncertain, although there is clearly a legal and regulatory risk if marketing becomes so targeted that it essentially amounts to the de facto application of gender as an underwriting factor. 

Considerations for insurers

As the legislative and regulatory landscape and AI continues to evolve, insurance companies must stay informed and up to date to avoid inadvertently breaching laws or regulations.

Marketing communications must not contain anything likely to cause harm to people with protected characteristics. However, insurers do target products at people with particular protected characteristics.

Targeted marketing may produce outputs which have discriminatory effects on people because of the use of protected characteristics, such as gender. As targeted marketing becomes increasingly effective and hyper personalised due to AI, the treatment of certain groups will need to be closely monitored. 

Firms are required to embed treating customers fairly into their corporate strategy and build it into their firms’ culture and day to day operations – the FCA expects to see fair outcomes for consumers. This means addressing the fair treatment of consumers throughout the product life-cycle – this includes marketing and promoting the product.

The Equality Act 2010 (“EA 2010”) gives individuals protection from direct and indirect discrimination, whether generated by human or an automated decision-making system (or a combination). Demonstrating that a system using AI is not unlawfully discriminatory under the EA 2010 is complex. Insurers are advised to carefully consider the implications of AI in their targeted marketing of insurance products.