AI governance: Ensuring a trusted and financially inclusive insurance sector

Contribution to the Eurofi Magazine - February 2022

Without question, artificial intelligence – or AI – plays a vital role in Europe’s digital transformation and is already having – and will continue to have – considerable impact in all areas of financial services.

This is certainly true for the insurance sector, where data processing and mathematical models are at the core of any insurer’s business and are used to determine underwriting decisions, pricing policies, settle claims, as well as to prevent fraud. For example, the use of historical customer data and survey data is used to inform new products at the product development stage and micro segmentation and personalised pricing based on non-risk individual behavioural data to estimate price elasticity and churn propensity at the pricing and underwriting stage.  Other examples virtual assistants and chatbots that use natural language processing communicate with consumers as part of the sales or customer service process, and the automated segmentation of claims by type and complexity at the claims management stage.

In a thematic review published by the European Insurance and Occupational Pensions Authority (EIOPA) in 2019, some 30% of respondents indicated that they were already actively using AI in their business, with a further 25% at proof-of-concept stage.  Given the impetus that the COVID-19 pandemic has given to the digital transformation, the figures will have certainly increased during the last two years.

This is no bad thing. The use of AI in the insurance sector benefits both providers and consumers. Providers can expect cost efficiencies across the entire insurance value chain as AI improves prediction, accuracy and automation. Consumers also see benefits: A wider choice and better tailored products. In addition, they may notice the cost efficiencies from providers transformed into more competitively-priced products.

It is essential to ensure, however, that AI does not lead to the exclusion different groups of people. This is particularly relevant given the trend towards increasingly data-driven business models, where the need for high quality data that is free from bias becomes all the more important so as to avoid discriminatory outcomes of AI systems, often disproportionately affecting vulnerable groups. 

Such issues have been highlighted by EIOPA’s consultative expert group on digital ethics in insurance. The principles developed by this multidisciplinary stakeholder group take into account the specificities of the insurance sector and lay down the key governance pillars for the ethical and trustworthy use of AI in insurance. The group’s report on AI governance principles provides guidance to insurers on how to implement key principles in practice throughout the lifecycle of an AI application, which at the same time needs to be adapted to concrete AI use cases. 

Transparency and explainability of AI algorithms play a key role here, especially in customer-facing AI applications. Indeed insurers must be able to meaningfully explain how their algorithms work and be accountable for their systems, enabling consumers to have access to adequate redress mechanisms.

This is where regulation and supervision can help to ensure good consumer outcomes.

Legislation also needs to adapt to the digital age. In this regard, the European Commission’s legislative proposal for an Artificial Intelligence Act will provide a legal framework for the use of AI in the European Union. EIOPA supports the Commission’s risk-based approach reflected in the AI Act. Indeed, not all AI systems pose the same opportunities and risks and hence the need for proportionality.

In itself, the insurance sector is already highly regulated. Sector-specific regulatory frameworks, such as the Solvency II Directive and the Insurance Distribution Directive, already cover the use of AI. In addition, cross-sectoral frameworks, such as the General Data Protection Regulation (GDPR) provide a sound basis for AI in the insurance sector. Further regulation should also take into account AI use cases in insurance.

Finally, the fast pace of technological development worldwide, coupled with growing global interconnectedness underlines why international coordination is vital. EIOPA works closely with colleagues at international level, in particular in the International Association of Insurance Supervisors (IAIS) to ensure a common coordinated approach at global level.

To conclude, AI is transforming the way people live, touching all aspects of daily life from education to the workplace to the home. This is why its use must be ethical, fair and without bias.

Developing a sound AI governance framework is therefore essential for a well-functioning, trusted and financially inclusive insurance sector and EIOPA will continue its work in this field to ensure good outcomes for policyholders.