Artificial intelligence in insurance: Unmasked

Envelop Risk’s Paul Guthrie on the evolution of AI’s role in the insurance sector.

The history of AI parallels the history of computing itself. Alan Turing’s seminal 1950 paper “Computing Machinery and Intelligence” explicitly discusses how to build machines and test their ‘intelligence’ – introducing the now famous “Imitation Game”.

The term and concept of AI as it’s used today is similarly ancient by computing standards; it was first applied and adopted after the (famous in maths and computing circles) Dartmouth Summer Research Project on Artificial Intelligence in 1955 in Hanover, New Hampshire. The six-week event defined the term and a cluster of component technical approaches such as machine learning, natural language programming, neural networks and other related areas that were included under the AI umbrella.

AI has ascended to the main stage this century. It is used extensively across governments and industries, becoming nearly synonymous with any kind of analytics using large amounts of data that can be processed without direct human intervention.

In 2021, New Vantage Partners reported that 99 percent of Fortune 1000 companies were using AI, with 65 percent investing over $50mn annually. And in 2023, it reported that 91.9 percent of those businesses had reported measurable business value from those investments, up from 48.4 percent in 2017.

Examples of businesses using AI abound, and most of us have been relying on it every day for some time. Google and Apple use it in their maps applications; Netflix recommends shows it thinks you’ll like; Amazon suggests products for you to consider based on previous shopping behaviour.

In healthcare settings, large companies like Siemens and General Electric use AI-based tools to speed up and improve diagnostics. In finance, AI is core to the strategies of venerable firms like Renaissance Technologies, whose famous Medallion fund innovated its use and built the best record in investing history.

Most recently, AI has revealed a new aspect of itself, with the launch and public experience of large language models such as ChatGPT 4.0, with many applications still emerging.

The point is – AI is not a fad. It is a core element of the general evolution of computing power and in many cases, AI is simply how maths is done.

Analysing six billion rows

One of the primary benefits of AI in insurance is the ability to design much more complex, data-driven, subtle, and ultimately more reliable models of real-world activity. The digital world is awash in data and the volume, velocity and interconnectedness of available data is ever-increasing.

In the past, without the resources to analyse data at scale, modelling teams would simplify what they were studying by using theoretical assumptions about the nature of the risk that vastly reduced the mathematics required. The accuracy of these models was then contingent on non-modelled assumptions. With AI tools that make analytics reliable, it is possible to assess much more complex systems, and the more data-driven models can be far more robust to changes in the underlying risk.

To illustrate via use case: Envelop Risk, a leading cyber (re)insurance firm, processes firmographic, economic and technical data on companies from across the world, combined with comprehensive claims data. It then adds comprehensive intelligence on cyber threat actors – their tactics, techniques, threat vectors and ongoing evolution.

The output of the in-house developed AI-driven algorithms creates a wealth of granular information that can provide underwriting teams with forensic detail, to enhance their decision making. The raw underlying data, would, if using the standard industry tools, require a six billion row Excel spreadsheet; a scroll length that would wrap around Earth’s equator roughly 1.5 times.

AI makes the creation of those six billion rows both possible, and the resulting insight useful.

The algorithms used in AI are a set of instructions to run certain kinds of operations across a dataset – such as regression or statistics – to calculate the relationships between variables and combinations of variables (or features) within the data. The algorithm will complete millions of calculations, along the way making programmed decisions on how to adjust the analysis based on interim results. The end effect has sometimes been compared to having “infinite interns”.

While AI is an evolution in the way insurers analyse data, the contribution of AI-based tools to the insurance industry is huge: it helps organisations work more quickly, with greater accuracy and validation, and – most importantly – utilise more data and handle the complexity that comes with an increasingly digital and interconnected world.

Insurance industry antagonist

As discussed above, AI delivers a tool which can learn and adapt to enable rapid analysis of complex data, and help data scientists, actuaries and underwriters determine their exposure and PMLs. On the claims side, AI has proven useful to analyse events leading up to losses to find commonalities that can be applied to future scenarios.

In a typical scenario, imagine that a well-known healthtech company suffers a ransomware attack. Over the course of the attack, the hackers encrypt the company’s systems, rendering them useless. The company’s customers are unable to upload data from their healthtech devices to the app and can’t access their accounts or stats. The hackers demand $10mn to decrypt the systems.

Following a scenario like the above, AI is critical to post-attack mitigation. AI enables the unique fingerprint of such events to be encoded. This encoding allows the fingerprint to be compared to other events that have been encoded in a similar way. It allows experience, root causes and mitigations for other close events to be discovered quickly and does so much faster and more reliably than by humans alone. It enables scenario analysis, forecasts of market losses and estimates of exposure.

AI will continue to evolve in its application and support for actuaries and underwriters. The industry owes it to their profession to cut through the glitz and hype that can surround AI and show leadership in how these tools can deliver value to their customers, their business, and their capital partners.