“Alexa, can you reorder toothpaste, get bottled water, and purchase a 20-year, $500,000 term life insurance policy?”
OK, we’re not there yet, but there has been a significant evolution in the application of artificial intelligence (AI) and machine learning within the life insurance industry.
It’s a natural fit for the capabilities of AI because of the large, complex data sets with nuanced relationships, years of historical data, and a unique sales process in need of a facelift.
AI is most commonly seen through the lens of natural language processing (NLP) — NLP takes over when someone asks Alexa for insurance quotes, interacts with social media chatbots, or even files an insurance claim.
In the pre-purchase education phase, AI bots could be used to help people understand their insurance needs, answer questions about their financial situation, and help customers continue with confidence down the path to purchase. It must be highly sophisticated and personalized to be truly useful, however, or else customers end up with “Sorry, I don’t understand that” responses.
There’s another significant opportunity for AI that takes an adaptive approach to the insurance-buying experience: tailoring the tone and purchasing journey based on specific customer profiles and inputs, which would ultimately remove the need for irrelevant questions and steps.
Machine-based algorithmic underwriting
As more data and experiences come in, machine learning technologies can iterate over permutations to find subtle patterns and relationships between data points that are only apparent after it acquires a greater population of applicants. It can go beyond human analysis to uncover intricacies that most people would miss.
This machine-based process gives life insurance customers added layers of value for the information they hand over during the application process. For example, it enables providing an instant decision on coverage and offering more competitive pricing due to higher accuracy and therefore less risk.
There are limitations, for now, with machine-based underwriting. The machine learning is used where it can come to a confident decision based on the data inputs and underwriting rules it receives. For more complex cases, or until it has learned from enough scenarios, the machine is programmed to know when the analysis should be handed off to a human for a more thorough review.
When a manual review is required, the machine can narrow down the details in a structured manner to enable the underwriter to focus specifically on what’s of interest.
Data: The foundation of life insurance
To understand machine learning in life insurance, you must consider the data needed to make a decision on coverage and confirm it’s correct. This is one of the most complex data sets to analyze and iterate from because it takes up to 30 years to see the result of an underwriting decision. There are two main categories of data used for machine learning in life insurance: applicant information and external data sources.
Valuable insights about a customer are gained in the application process. This is where machine learning is used to compare a person’s health history, lifestyle choices, occupations, and their subsequent risk to similar life insurance buyers.
To create a clear comparison, the model needs a history of underwriting decisions and what the outcomes were, third-party data sets, and underwriting rules to follow. For example, our algorithmic underwriting platform uses rules from 15 years of historical data and about 1 million applicants. These insights, combined with industry standard third-party data sources and applicant information, is how the model can come to a decision.
You might be thinking, “With daily activity, social and various other data sets available, why not explore new data sources?”
The answer: More data is not always a good thing, from both the customer and the technology’s standpoint.
If you’re asking for more data from a customer, then you must be providing more value for the information they are handing over. You also must ensure to ethically collect, analyze, and dispose of that data.
From the machine’s standpoint, too many data points with too few example scenarios can create too many variables for the machine to successfully determine what is “significant.”
The key is balancing the want for more data with increased precision and value.
Where we go from here
As more companies seek to deploy AI technologies, the focus should be customer value, first and foremost. If utilized correctly, machine learning can lessen the need to collect data in favor of only asking questions that are needed to determine mortality and ultimately come to a decision.
Although it might not be practical (yet) to order life insurance through Alexa, the idea of a fully machine-powered process for financially protecting your loved ones may not be that far away.
Todd Rodgers is the chief technology officer at Haven Life, a startup that helps improve the life insurance buying experience.