Medical information and data has grown exponentially in recent years, posing new challenges for life insurance underwriting. With often voluminous medical histories to assess risk, the process can take an inordinate amount of time. Applicants can end up frustrated, dropping out of the application process and seeking other alternatives, perhaps with competitors, in search of a quicker turnaround.

In response, insurers are turning to natural language processing – a more focused implementation of conversational AI – to assist with sifting through massive amounts of medical documentation to identify and even assess mortality risk. The benefits are manifold: not only does this result in an accelerated and accurate new business underwriting process, but it’s also a way to create a quality data set for improved predictive underwriting.

The Economics of AI

But many insurers are baffled by how to deploy AI economically. One way is to use software as a service (SaaS). For example, our Risk Profile Gateway was launched in the U.S. earlier this year to deliver an advanced AI solution for life insurance underwriting. This approach allows insurers to access advanced AI capability on a pay-as-you-go basis, significantly reducing capital investments and operating costs without compromising business results. 

As the leader of this venture, I engage with a variety of industry stakeholders, including direct carriers, reinsurers, underwriting experts and technology integrators. While all are excited about AI as a technology, my conversations suggest that many industry insiders need help developing a clearer understanding of the economics of the technology.

By framing AI opportunities in terms of their impact on demand, supply, pricing and cost variables, stakeholders can accurately assess AI’s near-term potential, long-term benefits and necessary trade-offs. Our discussions with a handful of direct carriers and reinsurers are helping them understand the trade-offs between elements such as AI accuracy vs. breadth of risk information extracted, and speed of risk profile extraction vs. breadth of insurance use cases that can be pursued.

These revelations have encouraged many of our clients to consider niche AI applications, such as using machine intelligence to contend with decline triaging, large case handling, seasonal spikes and analytical database creation. Each of these applications addresses unique economic levers, thereby strengthening the collective business case for AI-based intervention even further.  

Getting Beyond AI Fear, Uncertainty & Doubt

Anytime you introduce something as paradigm-breaking as AI, there is a need to address the uncertainty it introduces to the process to which it is applied. Basic principles of economics provide a well-established framework to model this uncertainty, and what it means for decision makers.

For example, underwriting inherently involves making predictions about something that could happen in the future based on what you know today. As Ajay Agrawal, Joshua Gans and Avi Goldfarb explain in their wonderful book Prediction Machines, AI at its core is all about making predictions – making smarter predictions based on training data that iteratively teaches the AI model to make even better predictions over time. 

When we switch from a model in which humans make the majority of predictions to one that increasingly relies on machines – and do so at scale – the price paid to arrive at a prediction can significantly drop. And when something becomes significantly less expensive over a very short time, we tend to use it more often.

From electricity to the internet, the history of technological progress is replete with examples of this simple economic principle at work – a steep price decline drives higher consumption, whereby the underlying technology becomes a utility and spawns radically different business models.

The use of AI in life insurance underwriting will follow the same curve. Tools, techniques and partners have emerged that can help life insurers reduce cost and the manual effort necessary to review lengthy medical records.

But this is just the first step. With significant scaling of inexpensive and more accurate prediction capability feeding off actual medical data, it may be necessary to revisit the current trade-off of looking at Fair Credit Reporting Act non-medical data (i.e., credit history or other database-driven data). It may also become possible to reassess mortality estimates based on near real-time data because the costs associated with converting that data into underwriting decisions may plummet, which will make the embrace of AI even more commercially viable. The sky is the limit as to how all this could evolve.

AI’s Illuminating Power

As the insurance industry looks to accelerate the underwriting process, AI will create new trade-offs on machine-powered accuracy vs. the accuracy of using non-medical data to drive predictions. It will be useful to keep in mind that as businesses switched from candle light to electric light bulbs, they not only benefited from a dramatic cost reduction for illuminating their factories and offices, but they also learned how to harness electricity for many different business uses.

Factoring in the right trade-offs on accuracy and breadth of use cases, AI will reduce life insurance underwriting costs in the near term. Longer term, it will also change the way the industry assesses mortality risk because the customer experience of sharing health data, and the level of effort needed to interpret that data, will both be transformed through AI.

Life insurers must look beyond near-term AI-driven cost arbitrage to reimagine the entire industry in the age of AI and intelligent, continuously learning algorithms. The types of policies offered, risks covered and benefits realized by customers will be well beyond what has been the industry norm.

As you assess AI’s fit in your business, there are many lessons to take from life insurance. It will be essential to identify both use cases with tangible benefits and a partner that can help you navigate not only the technology but also the business implications. 

The author would like to thank Associate Director Sreenivas Rangamani and Senior Director Michael Wilson in Cognizant’s Insurance Practice, as well as Tom McCarthy, Senior Underwriting Consultant, for their contributions to this post.

Chandan Gokhale

Chandan Gokhale

Chandan Gokhale is Startup Leader of Cognizant Risk Profile Gateway, a Cognizant Accelerator Venture. His focus is on building a clinical health data clearinghouse... Read more