The use and impact of artificial intelligence (AI) in insurance has significant implications for both insurers and policyholders. For insurers, a growing number are turning to AI to assist with modelling and the placement of risk. For policyholders, however, who need appropriate cover in place, AI has become a risk in its own right.
How will developments in AI affect the insurance industry?
Whilst AI has long been used within the insurance industry, the impact of COVID-19 has resulted in increased digitisation for insurance companies. This means they are well-positioned to embrace developments in AI over the coming years.
Data collection occurs on a widespread commercial basis, with personal data being shared in ways of which many of us will be unaware. Open-source protocols will increasingly be required to share data safely across different industries. Developments in regulation and cybersecurity will be necessary to negate and mitigate against potential risks. For example, information from home security systems / devices could be shared with insurance companies to assist with a more personalised home insurance risk assessment.
Consumer data is already collected by the majority of companies. Provision of such data can be used by insurers to model risk and assist insurers with a deeper understanding of individual policyholders' risk profiles. For example, modern cars generally have an inbuilt computer system which could assist insurers with pricing car insurance, or smart watches / fitness trackers could be used to price health insurance on a more individual basis.
Cognitive technology and machine learning
Cognitive technology imitates the human brain's ability to learn. According to the International Data Corporation, spending on cognitive and AI systems was expected to reach US$ 77.6 billion in 2022 with a significant amount of that investment directed to conversational AI applications such as chatbots and deep learning and machine learning applications. Use and application of such technology will likely become a standard approach for processing the complex data streams that will be generated by insurance products which provide cover in respect of aspects of an individual’s behaviour and activities. The increased commercialisation in this area will result in insurers having access to models that are constantly learning and adapting to the world around them. Predictive data will enable insurers to utilise past and current behaviours to understand the evolution of risks, customer experiences and expectations and to enable innovation through new products. It will allow insurers to carry out more tailored placement of insurance, with less active input required from the policyholder, broker or insurer. Smart contracts enabled by Blockchain could also instantaneously authorise premium payments and claim settlements.
How will coverage need to change to address AI-related risks?
Robotics are now used in a wide range of industries and will be continue to be deployed at an ever increasing rate. Examples include:
- Manufacturing industries often rely on robots to form a significant part of their production line;
- Military use in times of conflict;
- 3D printing is becoming increasingly common resulting in robotically manufactured products;
- Retail – reports that Amazon relies on 45,000+ drones to make deliveries and "augmented reality" in the retail sector show how different sectors are investing in and developing AI to increase revenues;
- Vehicles are increasingly being produced with a number of autonomous features and (whilst driver control is still a legal requirement) self-driving cars are already a reality; and
- Surgical procedures involving extreme precision using robotic arms and other equipment.
There has been an inevitable and significant increase in tech start-ups which create and maintain AI.
These companies may start off with little regulation but then grow exponentially, potentially on a global scale. Often regulation and insurance risk analysis is slow to catch up with the growth and needs of these companies.
Policyholders (and their brokers) must ensure that adequate insurance is effected, and that there is an ongoing dialogue with the relevant insurers to discuss changes in risk or additional cover required as companies grow and develop.
Cyber insurance has been available for years but the scope of cover has been limited. Whilst cyber insurance remains a growing market, it needs to develop further in order to fully service corporates and to offer meaningful protection against exposure to potential data breaches. This is particularly important in circumstances where such breaches could be significant as a result of increased data storage due to AI / technological advances. Insurers need to work more closely with their policyholders to fully assess their risks and needs. It may also be that increased preventative measures are required, such as particular security systems or protocols/crisis response in the event of a breach, to meet policy requirements for cover.
Linked to the issue above is the global impact of AI and cyber. Policies tend to contain clauses specifying geographic limits, potentially excluding cover for losses in certain territories. In circumstances where AI losses may be remote, or the source of the loss is not readily identifiable, insurers will have to respond to global risks. This may lead to difficulties in identifying and properly pricing increased risks unless there is more of a "deep-dive" at the placement/renewal stages.
Whilst there are a number of AI products which will assist the insurance industry with assessing and pricing risk and tailoring products to consumers, AI will also inevitably lead to new and increased exposures for policyholders. Policyholders should take time to understand the scope of their cyber cover, discuss the available protections with their brokers and put in place procedures to mitigate risk of cyber claims. In a world where AI continues to develop quickly, policyholders need to properly consider the cover they are buying and whether it does in fact offer them the protection they need.