Editor's Note: This is part one of a two-part series examining the data challenges insurers face and the impact of AI as it is incorporated across their operations.
As carriers make the transition to incorporate AI across their operations, data stands at the heart of every integration. Insurance carriers have a plethora of information and the format it's in (handwritten notes vs. antiquated programs vs. Excel spreadsheets vs. a current digital database) and where it resides often affects its ability to be integrated into various AI models.
Internal data often involves information gathered by insurers such as policyholder demographics, coverage risks, pricing, claims details, payment history, health records, sales data and a wide range of financial information.
External data generally comes from sources outside of the carriers such as hazard and risk data for flooding or wildfires, historical and real-time weather data, aerial imagery, health records or driving data from telematics, valuations for property values, geospatial data for elevations, proximity to water sources, a range of public records, and even social media insights.
"
Prashat Hinge, chief information and transformation officer at MSIG USA concurs, adding that carriers have vast amounts of data related to policies, demographics, exposures, building risks and more, but they often struggle with organizing it and using it consistently. Internal data silos also highlight the value of being able to effectively combine internal and external data. "You can have good data, but it is hard to organize and use it. They need to create a data model that helps insurers access the data to make policy decisions."
Determining the value in data
What types of data are valuable to an insurer often depends on the department and areas involved. "Actuarial teams care deeply about historical loss performance and exposure data," shares Jennifer Linton, CEO of Fenris. "Underwriters want current risk context that helps them to assess eligibility and appetite alignment. Distribution teams want to understand whether a prospect is likely to bind, retain or fall out of appetite before time is invested. What we are seeing across the market is a push to connect those perspectives earlier in the workflow."
Adam Pichon, senior vice president of Global Analytics for LexisNexis Risk Solutions, explains that data needs also vary across different lines of insurance such as auto, home, commercial liability and life insurance. Within those lines there is also a need for different types of data related to general business trends, marketing, pricing, underwriting and claims handling.
"Insurers have performance data on their policies such as, 'we sold a policy and that had a claim'. This determines future pricing," he says. "They also use data that comes directly from customers to predict risk such as age, where you live, type of car or house."
"What data they need depends on the risk or area that is being written," shares Chris Rice, vice president of strategic business intelligence for LexisNexis Risk Solutions. "Small carriers used to share data to have a better idea of their risks and to help price them better."
Pichon explains that most of the third-party data insurers use is accurate and precise because it is regulated, such as credit reports. "They have to comply with the law." However, he also finds that some information about properties may not be completely accurate or can be incomplete, making it necessary to find data from other sources.
Sometimes the veracity of the data from policyholders could also be questionable, so external data can be valuable for information such as the price of a commercial building, home or vehicle.
Scot Barton, chief product officer at Carpe Data, says carriers can verify information via entities like "Google Earth and through a number of third-party data providers like Lexus Nexus and Verisk." He shares that while video may be too expensive to use at scale, insurance companies can look at photos online to gather more information for underwriting and claims.
Finding multiple data sources
Bob Black, National Property Practice Leader for Amwins, agrees that there are a number of valuable tools that can provide important data. "All day long, Google Maps is getting used extensively. There's a product that's sort of old, but it's effective, called Microsoft Map Point. There's a product one of the carriers told me about called easymapmaker.com…and something called SpatialKey." Each of these programs help insurance professionals create custom maps to provide a better understanding of the risks for the areas being underwritten.
Other outside data sources can include federal and state level data, permits and even OSHA, says Hinge. "To write a risk, you can use satellite images, building demographics and you can get very precise. However, it's not a data problem, it's an integration problem. You need to investigate the data for a specific problem and Co-Pilot can help with the right prompts, but you need the tight level of integration to get the information you need."
"Everyone chooses to underwrite their risk differently," says Black, and in terms of AI use, "some are on the very front end, if not the beginning, and then some are moving down the path of being able to take a submission that I send and automatically pull it into their system and have the system tee something up for the underwriter before anybody even touches it."
Whatever the data source – internal or external – carriers and brokers still face similar issues. "Access is rarely the limiting factor, orchestration is," finds Bourland. "The complexity comes from validating the data, ensuring compliance, and integrating it into core systems in a way that improves decisioning without slowing down the workflow."
Part two of this series will focus on how artificial intelligence is affecting the data process and how carriers are protecting the information that goes into AI algorithms.
