Gen AI makes data that much more valuable to insurers
Third in a series on how the insurance industry now sees Gen AI and its potential.
Data can be a big key to support the application of Gen AI to risk management. With that in mind, the scope of data relevant to covering commercial enterprises can be overwhelming.
Rima Safari, partner, PWC.
Picasa
“In group insurance, there is a lot of data to ingest,” said Rima Safari, a partner in the insurance practice at PWC. “Whether it is a commercial lines insurance carrier trying to insure a marine shipping company, and they have to look at all the vessels involved, or a group insurance carrier looking at insuring an employer, and now you’re looking at the 50,000 plus employees and their demographics to be able to drive insurance.”
Being able to use Gen AI also depends on being able to clean up data and get a “single source of truth,” a common IT concept meaning one single reference point for all of a company’s data, as Safari explained. Using Gen AI for insurance underwriting and claims will depend on how data is input and summarized, she said.
Gen AI’s ability to handle data is powerful because it can ask for and access specific information, according to Monica Minkel, vice president and executive risk enterprise leader at Holmes Murphy, an insurance brokerage and risk management firm based in Waukee, Iowa.
Monica Minkel, vice president and executive risk enterprise leader at Holmes Murphy.
“Think about the filing cabinet that’s locked behind the desk of the CEO. Probably really sensitive information in there, but when it’s all online, when it’s all accessible electronically, and then you put in something like an AI tool that can scan, can analyze, can access all of this data, someone who may be able to ask a question of that AI tool may find themselves accessing data that they shouldn’t be able to access,” she said. “So it’s critical that companies are really mindful about permission management, and they don’t let information get into the wrong hands.”
Applying Gen AI in this way makes it possible to “ask a document a question – even ask an insurance policy a question,” Minkel added. “It’s interesting to think about asking a document a question, but whenever we think about how we might read a document, we’re looking for information.”
To that end, Nationwide has an enterprise data department handling data governance and management. That department is part of an ecosystem that includes enterprise risk management, internal audit, privacy and security teams.
Todd Lukens, senior vice president and chief technology and information security officer, Nationwide.
“We have it looking at quality, sufficiency and protection, and under that protection are other things, like privacy and ethics,” said Todd Lukens, senior vice president and chief technology and information security officer at the carrier. “We want to make sure we are using the data in an ethical way. Our compliance, legal compliance team and our chief ethics officer is involved in those conversations as well, so we’re looking at it from all different angles.”
With this data department in place, Nationwide protects its data set from being used in broader models or by others’ use of large language models (LLMs). “We think about it from a risk perspective,” Lukens said. “Those are some of the primary things we always anchor on as we look at the utilization of Gen AI capabilities.”
Applying Gen AI to insurance data makes it more valuable, and therefore more important to protect, as Minkel of Holmes Murphy stated.
“Regardless of where that information is coming from, we need to be very thoughtful about how we push it back out, because we are ultimately responsible for what we do,” she said.
