Transforming insurance with generative AI

Artificial intelligence (AI) is rapidly transforming many industries, and insurance is no exception. Its impact is only expected to grow as its capabilities expand – “by an order of magnitude next year”, as Elon Musk recently said.

AI is being used to automate tasks, improve risk assessment and develop new insurance products. It can enhance customer experience, increasing conversion rates and strengthen loyalty. And within insurance company operations, it can also automate routine processes and expand productivity. By delivering these benefits, AI will inevitably generate augmented profit levels. Insurance companies must engage with this technology if they are to thrive in the future.

Market insights and acquiring new customers

While a good deal of recent media coverage has centred around new “generative” AI models that can create text and images, such as ChatGPT and Midjourney, it’s important to recognise that there is another, longer-established set of AI models that work in a slightly different way. These are known as “discriminative” AI. This type of AI is highly efficient at identifying patterns in large data sets and is at the heart of the advanced data analytics that can reveal behavioural patterns and hidden demographic characteristics.

These tools help insurers know where to target their product development and marketing efforts. In particular, analytics driven by AI can identify whether a particular group has a low or high propensity to buy insurance, enabling targeted communications to be developed.

AI tools can also help with the creation of personalised communications and personalised insurance offers. AI-powered insurance companies can offer policies that are tailored to the specific needs of individual customers, with policies automatically written to their precise specifications.

In addition, AI enables them to offer policies that are highly price-competitive for individual customers: actuaries can generate insight from a variety of very different data such as driving records, travel patterns and social media activity, and use this to offer policy prices that accurately reflect the level of risk that an individual customer represents.

Developing new products and services

AI is also facilitating the development of new types of insurance such as parametric insurance and on-demand insurance. Parametric policies pay out a fixed amount if a specific event, such as a hurricane or earthquake, happens. AI is crucial for calculating the probabilities of these events, and thus ensuring that the insurance can be profitable.

On-demand insurance policies allow customers to purchase insurance coverage for a specific period of time, such as when they are on holiday or when they are hiring a car. Where the insured period is short, it is harder to calculate the risk (unless there are large numbers of policies that have been sold) and so again AI can help to ensure profitable business.

AI is also used for pay-as-you-go insurance policies, where customers pay for their insurance based on usage rather than paying a flat monthly or annual premium. Again, risk levels can be hard to calculate manually but AI can enable a more sophisticated approach to underwriting because data from a variety of sources, including unstructured data such as from social media and customer feedback, can be analysed in a cost-effective manner.

Increasing efficiency

Insurance can be a labour-intensive activity with complex risk evaluation and claims adjustment processes. AI can improve efficiency in many ways, for example summarising large quantities of content gathered during a claim, including call transcripts, agent notes and legal or medical reports. Much of this work is tedious but delegating it to an AI tool frees up insurance company employees so they can generate extra value by keeping customers satisfied and upselling or cross-selling other products.

AI can also decrease operational risks, with automated compliance monitoring and fraud detection. It can even be used to create training materials that keep staff up to date with the latest regulations.

Managing customers

The personalisation of policies and communications was mentioned earlier. But AI can take insurance companies further along this road, helping clients mitigate risks or even eliminate them before they happen.

Unless they’re trying to do it fraudulently, most people who take out insurance do not want to find themselves having to make a claim. Offering an AI-powered water leak sensor to policy households could mean that the householders are able to detect potential leaks before they become a problem, saving themselves heartache and the insurance company the cost of a claim. This type of proactive customer management benefits both the customer and the insurance company.

AI can make the process of dealing with an insurance company easier for customers. Using AI to automate the claims process makes it faster and easier for customers. For example, AI-powered chatbots can help customers file claims as well as answering questions about the claims process. In some cases, perhaps where they have had an accident that was their fault, people prefer talking to a machine rather than a human who they feel might be critical. For this to work effectively, customers must know when they are talking to a machine and have the option of moving to a human when they wish. 

Achieving success with AI

While AI is very powerful, care is needed to ensure it becomes the solution and not another problem. Insurance companies will not always find it easy to succeed in using it effectively, and there are risks that need to be navigated.

First, there is a need for knowledge and for people with the right experience and mindset. To handle AI, businesses need to establish a multidisciplinary team across different functions including IT, data analysis, compliance and communication. And while ambition is important, for most companies it will be sensible to develop the requisite expertise by beginning with simple and low-risk use-cases, such as the generation of social media posts and gradually fine-tuning models based on experience and measurement of outcomes.

There is also a need to understand the risks involved. AI can be used to identify risks, in cyber-security for instance, that humans fail to spot. But the technology itself carries risks, including problems with lawful IP usage, corporate-level reputation damage caused by bias, and information security risks.

To mitigate these risks, strong controls are needed, with appropriate compliance frameworks in place and always with accountability (as opposed to responsibility) allocated at a senior level. There may well be risks that insurers are unfamiliar with. Businesses need to update their risk management protocols to take account of them, and ensure employees are sufficiently aware of what can go wrong.

In addition, to avoid ethical problems such as accidental discrimination against certain groups, there is a need for human oversight and a constant and diligent search for inequalities. The reliability of AI-generated material and outputs can vary and, it must be accepted, can sometimes be very poor. The degree to which important decisions, such as whether to turn down someone for insurance, can be delegated to AI should be considered extremely carefully: there will always be a need for human oversight where decisions that could negatively impact people are concerned.

Used responsibly, AI has the power to transform insurance by increasing operational effectiveness and strengthening customer engagement. But insurers must proceed carefully if they hope to use AI to deliver trustworthy services. Strong guardrails will be needed to manage the risks associated with the technology. However, companies that adopt responsible AI practices will be well positioned to seize new and exciting opportunities as they arise in a rapidly evolving AI landscape.