The idea of artificial intelligence (AI) has fascinated us for decades.
The word “robot” seems to have originated from a play called “Rossum's Universal Robots” written by Czech playwright Karel Capek in 1920. Robots were not machines, but creatures that could be mistaken for humans.
The concept of more advanced types of AI emerged after World War II, when mathematician Alan Turing devised the Turing test in 1950. The Turing test is a test to assess a machine's ability to exhibit intelligent behavior indistinguishable from humans.
AI could have a big impact on underwriting
As AI becomes more of a reality, its portrayal in pop culture has become darker, with films like Terminator 2 depicting it as leading to nuclear war.
Exponential advances in technology and computing power over the past two decades have made true AI a reality. But with every new tool and system seemingly labeled AI, what does the term actually mean? What can AI do in financial services today? And where is it taking us?
Self-directed learning
What AI is not is a set of pre-programmed outcomes based on user choices. Imagine a vending machine. It operates based on user requests, but it is not AI, even though some fintech providers call it AI.
True AI involves learning, developing and growing, which computers can achieve on their own. ChatGPT is one of the best-known examples of proper AI, learning and improving the more you use it.
In financial services, this type of AI can play a major role in carrying out some of the time-consuming tasks that need to be performed, from administrative tasks like record-keeping and note-taking to initial customer triage, making work much more efficient.
Every new tool or system seems to be labeled AI.
Our area of expertise is insurance, where AI can have a profound impact on underwriting and pricing: Our decade-long study of AI underwriting (to determine how it can improve upon long-established methods that often lead to unfavorable outcomes for customers) revealed that age and other historically determined claims patterns should not be used as the primary determinants of price and excess.
Through large-scale analysis of claims data and cancellation patterns, we discovered that there are better predictors of risk and anti-selection. A much more accurate way to predict claims propensity and duration is to leverage global data and use state-of-the-art large-scale language models to analyze other lifestyle and behavioral factors through AI. This can include everything from information about an individual's occupation, to the mobile device they use, to the number of leisure trips they take per year.
By combining multiple layers of real-time data from global data agencies and leveraging detailed regression and correlation models, we found that some customers were deemed lower risk and therefore paid lower premiums than they would have under a traditional underwriting model.
We are testing this underwriting on accident, illness and unemployment insurance claims data and it is currently 80% accurate, but this will improve as the AI learns to analyse the data.
AI can play a major role in making our work much more efficient.
Ultimately, this type of AI-driven, smarter underwriting could be leveraged across financial services to match risk and price more accurately and quickly, creating a fairer, more efficient industry and giving insurers greater certainty that their policies are balanced.
There is no reason why this couldn’t work in future across all areas of financial services that require risk analysis, from mortgages and income protection to home insurance and even mortgage applications.
Kesh Tukaram is the co-founder of Best Insurance.
This article appeared in the July/August 2024 issue Mortgage Strategy.
If you would like to subscribe to our monthly print or digital magazine, please click here.
