Award-winning researcher Guandong Xu is a professor in the School of Computer Science and Data Science Institute at UTS. Photo by Andy Roberts
The underwriting process for life insurance is important but laborious for customer and insurer alike, as they pore over questions about work, life and medical history. But the application of artificial intelligence (AI) and data science to the process means underwriters at Zurich’s OnePath Life can now do in seconds what used to take hours – and customers no longer need to know their lateral epicondylitis from their tennis elbow.
This AI-data driven underwriting solution is a first of its kind.
— Peter Tilocca, Chief Underwriter, OnePath
Scientists from the Data Science Institute at the University of Technology Sydney (UTS) and OnePath Life underwriters combined their expertise to make the life insurance process easier, quicker and more affordable.
“This AI-data driven underwriting solution is a first of its kind and based on one of the largest sets of insurance data in the sector,” OnePath's Chief Underwriter Peter Tilocca says. “The insights gained through data mining enabled us to simplify the customer application and develop a ‘virtual underwriting’ tool.”
Today, thousands of OnePath underwriters are using this AI Virtual Underwriter, freeing them up to apply their specialist expertise to the more complex enquiries.
Time for change
Underwriting is the process by which an institution takes on financial risk for a fee – in this case insuring an individual’s life. It involves research to understand the degree of risk, to set a premium appropriate for taking on that risk.
The process hadn’t changed much in decades when OnePath, then owned by ANZ Banking Group, came together with data scientists at UTS in 2016.
With over 100 questions per application, and more than 1000 possible exclusions or loadings having to be manually assessed by underwriters, it could take up to a month to provide an offer to customers, OnePath’s Peter Tilocca says. Some customers gave up mid-application.
OnePath, a leading Australian insurer serving more than 2 million customers, wanted to simplify the process, for everyone’s sake.
It has been able to do this by harnessing the power of big data and machine learning to transform a very manual, process-oriented business into one powered by AI, data insights and decision-making tools.
We’ve shown how big data, natural language processing and predictive modelling can revolutionise a huge process like underwriting.
— Professor Guandong Xu, School of Computer Science
Connections revealed
“Using AI to simulate people’s decision-making process, we’ve shown how big data, natural language processing and predictive modelling can revolutionise a huge process like underwriting,” says Professor Guandong Xu, who led the UTS scientists working on the OnePath project.
The team used data science to explore 10 years' worth of data extracted from three core systems at OnePath – in all, tens of millions of rows of data. Advanced analytics and machine learning techniques uncovered connections between customer cohorts, underwriting questions and answers, and resulting insurance claims.
“We discovered that many questions don’t provide useful insight to determine whether an application should be accepted, have an exclusion or a loading, or be declined,” Prof Xu says. “We found that using just 10 per cent of the questions could achieve the same underwriting outcomes as using all the questions.”
A pilot project resulted in one section of an application being reduced from 32 questions to just seven, for instance, without any degradation of the insurance risk. “Essentially, we’re getting the same decisions but with a much better customer experience,” Peter Tilocca says.
In addition, the project means customers can now disclose medical information using everyday language – selecting ‘tennis elbow’ rather than that ‘lateral epicondylitis’ option, for instance – after the addition of about 1,000 aliases for medical conditions to the system’s master catalogue. This has significantly reduced the chance of customers diverting from pre-populated drop-down menus to the “other” text space, where free-form entries require the scrutiny of an underwriter.
The impact
The work has had clear outcomes for the business including:
- improved customer experience
- 30% faster completion rates
- higher straight-through processing and acceptance rates
- reduced underwriting referral triggers
- reduced annual operating costs
- automated quality assurance for all virtual and human underwriting
- high ratings from insurance adviser clients for the new system.
It’s award-winning work, with UTS and Zurich OnePath achieving global recognition at the 2020 Efma-Accenture Innovation in Insurance Awards, where Zurich was named Global Innovator and the project was recognised in the Workforce Transformation category. The UTS–OnePath partnership was named a Gold Disruptor in the 2019 ACS Digital Disruptors Awards, for service transformation, and secured three merit awards in the 2019 NSW iAwards.
Professor Xu says it shows how universities and industry can collaborate to achieve commercial outcomes in an agile way. “Bringing together different sectors for interaction and creativity produces real-world impact and results,” he says. “We’ve shown that Australia is capable of developing and translating world-leading technology to enable a company to build a globally competitive advantage.”
Peter Tilocca says underwriting is still a people business and a highly skilled underwriter is still worth their weight in gold. “They say underwriting is part art, part science,” he says. “I'm just trying to increase the science.”
Report/paper
Biddle R., Liu S., Tilocca P., Xu G. (2018) Automated Underwriting in Life Insurance: Predictions and Optimisation. In: Wang J., Cong G., Chen J., Qi J. (eds) Databases Theory and Applications. ADC 2018. Lecture Notes in Computer Science, vol 10837. Springer, Cham. https://doi.org/10.1007/978-3-319-92013-9_11
Research team
-
Professor, School of Computer Science
Faculty
- Faculty of Engineering and Information Technology
- Data Science Institute
Funded by
- OnePath Life
- Zurich Insurance Group (formerly ANZ Banking Group)
