The discussion around underwriting of the future happens a lot in our industry, with many looking towards technology and AI as the solution. This will no doubt play a crucial role and, as underwriting and claims professionals, some consideration needs to be given to how we can successfully embrace these changes and how these can benefit our customers.
Thinking back to when I first arrived in Singapore in 2017, I needed a bank account that could transact in both S$ and GBP, so a friend recommended one of the early digital-only banks. I will forever remember the registration experience – I expected it to be time consuming, onerous and requiring many documents and phone calls.
How wrong I was. It involved automatic reading of my documents and security checks where I recorded a video, reading the script, “My name is Andrew Doran and would like to open a bank account”.
Had the process been even a slight improvement on my expectations I would have been happy, but the reality blew me away and immediately made me think about how I should use this experience to influence my own role in underwriting and claims.
So, how should we approach both improving the experience for the customer and the understanding of the risk being taken?
AI and data analytics
When it comes to evolving the underwriting process, AI and data analytics continue to be the hot topics, with generally two schools of thought around how they can best be used.
Firstly, predictive underwriting – using data analytics techniques to determine the underwriting outcome from insights gained from both traditional and less traditional data, rather than through standard question and answer applications.
Or secondly, fully-automated underwriting with direct access to digital (health and non-health) information. Both have the intention of streamlining the underwriting process, reducing the number of questions that customers need to answer and improving understanding of the risk costs.
However, working within underwriting and claims for many years, it can be hard to turn my back completely on what already works well – the traditional application process with questions and answers.
The solution will almost certainly be a combination of all three.
I have at times imagined a future for underwriting that looks something like the below.
While this will be a direction taken by some, it is one of the most challenging approaches from a technical, regulatory and customer trust perspective.
- Technically it requires a complex solution involving digital health records, automated reading of medical information and automated cognitive reasoning to replicate the human underwriting decision-making process.
- Strict regulation around the use of medical information, whether Personal Data Protection Act (PDPA) in Singapore, Personal Information Protection Act (PIPA) in Korea or General Data Protection Regulation (GDPR) in Europe, would need to be adhered to.
- Building the customers’ trust that their most sensitive information is being used responsibly is of course central to creating a sustainable solution.
I am not saying that these challenges should not and would not be overcome but I consider the clear next best action as the coming together of traditional risk assessment augmented with elements of predictive underwriting using certain alternative data sources.
A work in progress
I see this as the long-term solution that will slowly evolve over time, at least for the mass markets.
Thinking for a moment about these ‘alternative data sources’, how exactly can the length of someone’s driveway predict whether they’ll have a heart attack in the future?
I have both asked and been asked this type of question many times. And linking driveway length to health is unusual but certainly doable.
There is a lot to be learned from more abstract data that may have correlations to health outcomes but with the increased focus on insurers more clearly explaining underwriting decisions to customers, I prefer to start by focusing on areas that can be more easily related to health, and work my way towards the less tangible data sources, such as driveway length.
Improvement through small iterations has always worked well for me. One of my previous bosses used to tell me, “Just because you can does not mean you should”. That has stuck with me – the message that trying to be too clever can often get in the way of taking a step forward.
Broadly speaking, the mix of underwriting outcomes leans heavily toward ‘standard rates’ decisions, at c.90%, whether it be in Singapore, Malaysia or the UK.
However, not everyone considered ‘standard’ carries the same level of risk and focusing on how data analysis can more accurately differentiate this group has clear advantages for both insurers and customers.
Whether allowing the extension of non-medical underwriting limits, offering a better price or reducing the number of underwriting questions. This is a much more manageable problem to tackle or at least to start with.
New insights can be valuable but it’s important to consider the context and any unintended consequences. I was asked this particular question by a senior ex-colleague, “Andrew, you report that your department pays more than 99% of claims, therefore should we simply remove the claims assessment process in favour of something more transactional and just pay 100% of claims?”
As you may expect, I did not move to paying 100% of claims without an assessment process. Although the data is compelling, the unintended consequences would have been disastrous.
But it did make me consider how we could better assess the 99%+ of claims that we did pay. How could we identify the types of claims that should be paid immediately and separate these from the claims that require a more robust process?
Thinking back to the customers
When people buy insurance, they hope they never need to use it, rather it gives them a safety net, should the worst happen. No one wants to claim.
During the underwriting process, we have obtained a lot of information about a person’s health – a detailed history of their past and present illnesses, possibly combined with doctors’ reports, blood tests and ECGs.
Moving forward, that will be supplemented further with other information, through wearables, implants or insights drawn from other predictive underwriting sources.
It is very much a one-way street with customers sharing lots of information with underwriters in return for acceptance of their application.
I believe that we would be missing a great opportunity if we embraced the advantages of data analytics and new technologies to improve the underwriting process, but left out the customer.
Instead, how can these advances be used not only to improve the underwriting process and our understanding of risk, but also to give meaningful health insights to customers? Giving them useful, straightforward information about their health risk which they can use to take any actions needed. That can only be a win-win for the industry and its customers.
As an underwriting and claims professional used to an approach that works well today, it can be overwhelming thinking about what might be possible using AI in our profession. Keep in mind what problem we are trying to solve – for customers and advisers – rather than on what is simply possible. It is important to have a vision but that should not get in the way of taking the first step. A
Mr Andrew Doran is head of underwriting and claims at Pacific Life Re Asia.