As ever, change is constant, but these days we all notice it more because the pace of change is quicker and still accelerating. In the 1990s one could look back and find that life and disability underwriting was little different from how it had been 50 years before. In fact, SelectX was formed getting on for 20 years ago with the aim of prodding the industry towards better ways of doing things and moving underwriting truly into the 21st century. Today, encouraging companies to innovate is largely unnecessary: they get the need. Instead, they need pushing onto the up-escalator of underwriting process revolution; miss the opportunity and you’ll die.
Drivers of change
So, what’s powering that escalator? Well, no prizes for guessing that it’s primarily technology. Although automated underwriting is an old concept, systems have got progressively smarter, and they and their associated modules have become extremely capable. Then there is the use of external databases, a) to aid initial risk triage (think identity verification and credit scores), b) to drive more accurate underwriting decisions or c) as a basis for predictive analytics to act as a substitute for part of the traditional underwriting process.
OK, all that is progressive but not exactly revolutionary. But use of advanced computing – machine learning, artificial intelligence, etc – is (at least for now). For example, these techniques can convert unstructured data into structured form – eg interpreting APSs, lab tests and other medical records – thus enabling information review and provisional decision-making without human involvement; and they are comparing the characteristics of a current case with those of previous ones and suggesting underwriting decisions based on the comparison (rather like ‘people like you also bought X, Y and Z’ that we are all familiar with when shopping on line). In time the machine will be making the final decision, and not just suggesting one for human approval.
While technology is the main driver, consumers are playing a part too. As we have observed before, they have come to expect the speed, simplicity, service and new forms of value that technology enables. And if they see a benefit, they will share their health information too; look at the ‘health partnership’ proposition from companies like Vitality. Our bet is that consumers will be willing to share more and more detailed personal health data – provided they trust the insurer and that doing so creates genuine value for them as a customer. There is growing acknowledgement that individuals own their health records and data, and those individuals will progressively feel more in control of it.
In terms of health information sharing, one naturally thinks of traditional health information. But according to AI specialists Lapetus, a simple selfie contains enough important biomarker information to drive much of the underwriting/pricing process. Is the snapshot a guide to the future and, indeed, the future itself?
And the point is?
So far we have talked about the ‘what’, but even more important is the ‘why’. The underwriting goals of insurers have to be:
- Making buying insurance as quick and easy as possible
- More appealing products
- Faster, more accurate and fairer underwriting
- More consistent underwriting
- Improved customer satisfaction, for better persistency and brand loyalty
- More underwriting efficiency and cost-effectiveness and at the same time improving risk management
- Better data for a better understanding of the business.
What might this ‘new underwriting’ enable? How might the world look for insurers, underwriters and consumers? The increasing power of automated underwriting will mean that this technology will be applied to a wider variety of risks, including more and even complex substandard ones. Arguably, there will be little need for human underwriters.
But will more or less complete automation really be the solution for everyone? One could envisage, for certain market segments, the emergence of more personal propositions in which individual risks are looked at and priced more carefully; this would likely involve a high degree of face-to-face adviser-client dealing.
Underwriting/risk assessment will become a central part of the customer proposition. Underwriting as part of product positioning is something we have believed in for a long time, but in future the scope to do so will be wider. Of course, Vitality kicked this off some years ago but there are other possibilities, not least furthering of the concept of an insured/insurer health partnership for mutual benefit.
What will be the role of reinsurers in future? Maybe they will dominate underwriting because of the data they routinely garner from multiple insurer clients and their expertise to analyse it, creating knowledge supremacy. But then reinsurers are not the only firms with data and the expertise to create value from it; there are also the specialists in this field who are pioneering data science, and they are nimble too. Maybe independent data science hubs will spring up to rival the reinsurers. A by-product of this would be the ability of carriers to maintain some independence from reinsurers, whom some believe hold too much power via their involvement with underwriting systems and accumulation of data.
Whichever way things go, could we see underwriting expertise gravitating towards specialists in risk data and ceasing to be a core competence of insurers? Note that the concept of the ‘virtual insurer’, in which the value chain is formed by a network of providers, has been a reality for some time.
Will we see fast-paced unbridled underwriting innovation, or will barriers slow the rate of change? Barriers or brakes are likely to come from two sources: consumers and regulators. There could be a consumer backlash if ‘revolutionised’ underwriting is considered to be unfair. While experience outcomes at a portfolio level might be fine, will ultra-automated underwriting differentiate enough between individual customers? Will it maintain equity between policyholders? Consumers need to understand how their terms were arrived at. If a decision is ‘right’ by the system but does not make sense to a human reviewing it, that decision is wrong.
And insurers’ ‘right to underwrite’ is increasingly being questioned, with allegations of discrimination and treating unfairly those disadvantaged by poor health. (Of course, underwriting is discriminatory, but with equity between policyholders and the avoidance of anti-selection as the primary goals. Discrimination can be viewed differently depending on where one is standing.) Raised more recently is the right to have medical conditions ‘forgotten’ about, maybe if there have been no symptoms and no need for treatment over a specified period – even though some conditions, for example cancer, have a long risk tail.
Both these issues sit at the point where consumer and regulator interests combine; either consumers raise and regulators become aware, or the regulator initiates the debate with consumers’ interests in mind. But either way, the outcome is a potential restriction on the scope of underwriting rather than the technical process.
To a degree, the success of risk selection making greater use of external data depends on the availability of that data. In Europe, for example, there is restricted scope because there are such tight data protection laws. Elsewhere, data privacy is more relaxed, at least for now; but you can bet that any trend of change will be in the direction of the European model.
And regulators are starting to look closely at how insurers use data. Even in the United States, traditionally a country of relative data freedom, state insurance commissioners are setting out codes of practice for insurers and threatening severe penalties for transgressors.
Finally, might entry into the new ‘machine-learnt’ world of underwriting be slowed by a scarcity of good data? In life insurance, underwriting produces plenty of acceptances but the mortality experience is usually very slow to emerge – thankfully, unlike motor vehicle accidents, deaths are usually pretty few and far between, so credible data takes time. OK, those creating underwriting philosophy have traditionally drawn heavily on population studies in their research, but how confident should chief underwriters and actuaries be in basing a whole new approach to risk selection and pricing on population data?
Maybe it will be a case of ‘suck it and see’, but if things go a bit awry, how will suitable adjustments be made? Will it be known what, in a complex algorithm, needs adjusting and by how much? Might equity between policyholders be at risk?
What might the job of underwriting look like in future?
Will there still be a need for underwriting manuals? Probably yes. After all, not all risks will be suited to automated assessment, and there are those highly tailored personal propositions to be catered for. Manuals will come closer to automated underwriting systems, sharing components like calculators. And they will be an integral part of underwriter ‘workbenches’ that act as data hubs, bringing together case management, external data feeds and all the tools needed to manage the underwriting operation. And these hubs will be the source of data to enable monitoring of performance every which way.
Whither underwriting as a profession? Well, there will be far fewer underwriters – the rise of the machines will take care of that. But the underwriting roles that remain will be more rewarding: less case underwriting (but more skilled assessment of the cases that remain), more time spent optimising engine performance, more management of the underwriting eco-system, and more interaction with data scientists and behavioural economists – and even with actuaries. Underwriting was always a great job for the people with a true aptitude for it, but maybe it is going to get even better.
We posed the question ‘Whither underwriting?’ Well, it won’t wither, it will just be very different.
*Whither: a rather archaic English word meaning ‘to where’. It is related to ‘hither’ (here) and ‘thither’ (there).