As a sector, insurance has not been good at the kind of deep data analytics we now see from machine learning. This is especially apparent in group insurance, which has traditionally lagged the investment provided to retail or individual insurance.

Give or take a few key terms, we are still, in effect, asking clients all the same questions we were more than two decades ago.

With the arrival in recent years of new technology, an odd contradiction has developed in insurance. Leaders in the industry know they must move to embrace artificial intelligence but they have been hesitant to commit to – and invest in – such systems.

We have seen some early adopters around the edges. In the UK, insurance group Neos Ventures offers competitively priced home insurance policies by installing smart monitoring devices in people’s house.

Perhaps the quirkiest standout is US group Lapetus Life Event Solutions, an insurer that offers cover based on a selfie alone. The company’s software can use the photo to build a basic risk profile based only on your face – taking into account factors such as if you are a smoker, for instance.

OnePath and UTS

At ANZ, the bank’s OnePath Life Insurance arm has had some early wins in AI-based collaboration with the University of Technology Sydney (UTS), primarily focusing on how AI and machine learning can lead to improvement in underwriting processes. The pilot program has allowed OnePath to slash the time and effort required to deliver what is ultimately a better service for both consumer and provider.

Through work with the UTS Advanced Analytics Institute, OnePath is exploring how modelling customer behaviour through machine learning, data science and probability can help add value and reduce the time it takes to secure a policy.

Previously, customers would receive a quote and were then invited to complete their medical history and a lengthy questionnaire. Only following this process – and potentially up to a month later – would they obtain confirmation or, in the worst-case scenario, news that they could not be covered at all.

Faster and better

In late 2017, a group of tradies in Victoria was invited to participate in a pilot testing a new predictive underwriting capability, which draws on big data and artificial intelligence to make the process faster, easier and more accurate.
The pilot was a success and led to a reduction in the application process from 32 questions to seven.

Using 10 years of insurance data, analysis by ANZ and UTS has made connections between customer segments, their questions and answers, and the resulting claims – allowing pinpoint correlations between responses to questions and claims.

This has enabled the OnePath team to pick out the most relevant medical questions in the underwriting process and discard the least relevant, reducing time, paperwork and cost.

This is potentially applicable to all forms of insurance; however, some of the biggest gains could be in the group insurance space.

Group insurance is typically offered to automatic acceptance levels but for members who want extra cover, the current application process can be even further reduced than it can be in retail insurance.

The advantage in group insurance is that members typically need to be at work on the date of commencement of the cover, thus reducing the anti-selection risk even further.

While this breakthrough in technology is a game-changer for underwriters, it is unlikely that AI will ever replace the underwriters themselves. The skills of an experienced underwriter will always be worth their weight in gold. Underwriting will always be part art, part science.

Join the discussion