Insurance data: Finding it, protecting it and the role of AI – Part 2

Editor's Note: This is part two of a two-part series examining the challenges insurers face and the impact of AI as it is incorporated across their operations. Part 1 explored data sources, who's using the information and its value to carriers. Part 2 looks at the impact of AI and the importance of protecting data assets.

There is no doubt that technology, particularly artificial intelligence, is changing every facet of the insurance process. From sales to underwriting to claims, AI, in particular, is answering questions, making recommendations, synthesizing large amounts of information for underwriters, flagging claims abnormalities and more.  

How is AI changing insurance and the data process?

The integration of AI across the insurance ecosystem comes with expectations that it will simplify or manage some tasks, highlight anomalies, and flag claims or underwriting scenarios that require more attention.

"AI is not typically discovering entirely new datasets," says Jay Bourland, CTO of Fenris. "Its strength is identifying patterns, inconsistencies, and predictive signals across structured information. It can flag anomalies, detect fraud indicators, and continuously recalibrate models as new data flows in. But AI depends on disciplined data management. It amplifies signal. It does not compensate for poor data hygiene."

"Carriers aren't using AI to price coverage," adds Adam Pichon, senior vice president of Global Analytics for LexisNexis Risk Solutions. "They aren't feeding risk data into AI because they don't want it to be exposed. However, it can be used to get data that comes from public websites and to summarize the data for an individual."

"It can also be used to improve the speed and efficiency of programming. It saves time and money and improves the speed of response," says Chris Rice, vice president of strategic business intelligence for LexisNexis Risk Solutions. "It can replace call centers and reduce human dependency in some areas."  

"Insurers have been using marketing AI for years," details Scot Barton, chief product officer at Carpe Data, "and it's solid for simple queries but not for automated decisioning. They are also cautious about what AI has access to, and it is embedded with controls when it's used. The speed and scale are impressive, particularly for coding and summations, but not for thinking decisions. You could ask a question three times and get three different answers, so there is still some inconsistency."

AI's ability to sort through a significant number of records and highlight their contents is impressive. "We used to take 2 million records and run a model overnight to pull out data," shares Pichon. "It could take weeks just to build one model. Now, we can run hundreds of millions of records through AI and have the results in minutes."

The value for insurers is the ability to have solid predictions for underwriting and risk management, provided the data they're working with is accurate. Inputting bad or incorrect data affects the veracity of the outcomes. "The insurance marketplace is highly competitive with small margins, and carriers only want good data that is predictive, because bad data doesn't help us," adds Pinchon.

AI is also helpful in flagging anomalies, whether it's related to fraudulent claims, identifying emerging risks or highlighting needed changes in existing underwriting models. "When you are analyzing millions of records, AI modeling can surface correlations and behavior patterns that are not obvious through manual review," says Bourland. "This is especially valuable in fraud detection, risk segmentation, and distribution optimization. However, models must be monitored and explainable. Insurance operates in a regulated environment. Performance and transparency both matter."

Even with these additional tools, human experience is still vital to interpreting everything from what an image may show to what trends are identified to ensure they are real and not a hallucination.

Protecting data is vital

A major concern for insurers and brokers is protecting the data they have collected. According to a study from Rocket Software, 69% of IT leaders cite data security as a top concern of modernizing their systems. Data laws are constantly changing and since insurance is regulated at the state level, what applies in one jurisdiction can be completely different in another. This is a particular concern with the introduction of AI, which may not be able to differentiate how compliance varies across the country, let alone around the globe.

"Carriers typically implement role-based access controls, encryption standards, audit logging, and segmented environments for development and testing," explains Jennifer Linton, CEO of Fenris. "There is also heightened scrutiny around external AI tools. Many insurers restrict sensitive data from public generative platforms and instead operate within controlled enterprise environments. Data protection is both a technical and regulatory obligation."

Prashant Hinge, chief information and transformation officer at MSIG USA, recommends simplifying a company's technology environment. "Cybersecurity has more technology and tools than any other area in insurance. Be more offensive than defensive and consider which business problems you're trying to solve. That's how you use AI."

Carpe Data's Barton sums up the challenges for insurers this way, "Carriers are drowning in data, but starving for decisions."

As insurers figure out ways to integrate AI safely into their data and technology stacks, years of information will be turned into valuable insights, enabling realistic pricing, personalized coverage and a better experience for all along the insurance continuum.