Powering the modern life insurance carrier with advanced AI and data

There continues to be significant opportunity for carriers who are able to make an efficient transition to a data-driven business model

Welcome back to our whitepaper series. If you read our introductory paper, you know that we are focused on how to use AI and data throughout the new business and underwriting processes. This white paper discusses using AI to build a data repository, create a baseline dataset, and conduct analyses to design new products in a shorter time frame and with a higher level of confidence in underwriting efficiency and product viability.

Building a data repository

The success of any new life insurance offering starts with product design, which in turn relies on data. Gathering and analyzing data has always been a major part of the insurance industry. That data collection can be split into two groups: data collected for immediate underwriting and data collected to evaluate underwriting experience. The traditional data collection process for underwriting a case begins with a records request going out to various vendors and putting together disparate pieces of information that can come digitally, via paper, or images. This information is laboriously collated to drive insights that can be used to set insurance prices. This same data is also aggregated to create larger data sets to develop better underwriting guidelines, shore up product assumptions, and used to validate new data sources that may be able to increase speed to market. Some of the data comes in a format that is ready to use for analysis, but the bulk of it, even digital records, is buried in unstructured documents such as PDFs and images. As underwriting evidence continues to evolve to include new sources of real time electronic data, there is the need to pair this data with the historically available information to generate insights and create credible data sets. Extracting usable, relevant information out of the historical data is where AI has its greatest immediate opportunity.

Using AI, machine learning, and natural language processing tools, carriers can convert their PDFs, images and paperbased business processes. AI-driven software solutions can uncover relevant data buried in existing documents with speed and efficiency – performing in weeks what would otherwise take months or years to accomplish. Furthermore, AI-driven software solutions can be used in underwriting to reduce cycle times and improve efficiency by taking incoming records and turning them into structured data that can be used in predictive analytics and rules engines to drive underwriting innovation with new rules and automatic decisions for some cases. For more complex cases, medical summarization and annotation can also be provided to speed up the decision process for cases that require full traditional underwriting.

Creating a baseline dataset

A detailed baseline dataset is necessary and impacts every stage involved in the product development process. Underwriters need to know how customer requirements can be combined and workflows modified to achieve the right risk classifications. Actuaries are looking at the combination of customer types needed to achieve specific financial outcomes. Marketing wants to know what kind of products appeal to the market. Having a baseline view of all aspects of the current offerings, customers and financial requirements puts the carrier in a stronger position to make better decisions in the product development process.

The accuracy of the data in the repository will drive the validity of the process all the way through claims. Its importance cannot be understated. Any AI or machine learning expert will tell you that models need to be trained for the context of the files they will be working with. Training can take thousands of cases to get to ideal levels of accuracy. How can this process be sped along, and what can be done in the interim to ensure accuracy targets are achieved until the model is fully trained?

Underwriters need to know how customer requirements can be combined and workflows modified to achieve the right risk classifications.

There has been a proliferation of AI and machine learning vendors in the past several years and that trend looks to continue. However, these solutions require more than technical capabilities to be successful in the near term. Models that have been designed and built for a specific industry and for specific use cases have taken advantage of existing information and terminology dictionaries to speed up the process. Even so, they still must deal with the variations of the unstructured data that they receive and the quality of the underlying sources of information. A strong baseline of experience and knowledge within an industry are important when developing models, but further refinement of the solution is still required by having a human in the loop as part of the process.

The human-in-the-loop concept requires a subject matter expert that understands the underlying business processes, the sources of data being fed into the extraction engine, and the uses of that data within the downstream processes. This person works with the source information and the extraction output to ensure that results are accurate and that the necessary adjustments are fed back into the AI or machine learning engine so they continue to evolve. As the system continues to train, the time needed for human intervention decreases until it evolves to a point that only random output verification is needed and ongoing oversight can be retired.

With this new data repository, a carrier’s product development team has a baseline dataset of its existing customers, their demographics, and medical information to use when designing new offerings. Creating a baseline of structured data allows the carrier to build key metrics from their historical book of business. These metrics could include high-level parameters such the percentage of customers per various underwriting classifications, or the rate of certain disease states in their current population. With the right baseline dataset, the parameters can be much more specific. Knowing this information is the foundation for anticipating impacts as a result of the changing underwriting evidence used in the decision-making process, such as using real-time data versions of electronic health records, clinical laboratory results or medical claims data rather than a traditional attending physician statement.

Designing the product and anticipating impacts

Working from the baseline, the product development team can use analytics techniques to forecast how changing any part of the product design could potentially affect results. Data can be quickly reviewed to determine anticipated underwriting outcomes based on input changes, such as requirement types and age or amount criteria. The goal of this exercise is to provide both information and confidence. Seeing the delta of change between the original underwriting decision and the new underwriting decision gives the product development team verifiable data to build a case for the success rate of a new product. From this more sophisticated model, predictions for fraudulent activity, policyholder behavior or other key factors can be developed in time. Most importantly, the development of leading indicators will allow the success rate to be tracked and adjusted as needed and closer to real time than waiting for early duration claims.

With a higher level of confidence in the data, products can be more quickly developed, implemented and marketed to the appropriate audiences. By using AI to review, analyze and adjust leading indicators, underwriting rules can be proactively adjusted to meet business outcomes and any issues can be addressed in real-time. Underwriting decisions will become increasingly automated, allowing underwriters to focus their work on high-value, complex cases. AI will reduce workloads, increase accuracy and get the customer results more quickly.

By using AI, machine learning and advanced analytics, carriers can reimagine how to design and test new insurance products and underwriting guidelines, bringing them to market more quickly, and with a much greater certainty of customer acceptance and financial success.

EXL Service offers AI-driven software solutions and professional expertise to life insurance clients to implement data extraction, data analytics and actuarial support so clients can more accurately define and build new products with anticipated financial results.

Read next in the series: Powering the modern life insurance carrier with advanced AI and data

Part 2: New product implementation—coming soon

Part 3: Product launch and distribution—coming soon

Part 4: Product monitoring and adjustments—coming soon


Written by:

Ammon Dixon
Head of New Business and Underwriting Solutions