With new products appearing on the market every month that feature artificial intelligence, and reports touting the technology’s potential impact on society, it’s easy for insurance professionals interested in adopting AI to feel they’ve had something of an information overload.
“What we’re seeing a lot of today is, I need machine learning because I’m reading about it everywhere, but there’s only certain points and areas within the business process and claims or underwriting or marketing that it actually makes sense,” said Mark Rusch, vice president of insurance at GoodData, which focuses on embedded and distributed analytics.
The company made moves into the insurance space recently with the launch of two analytical solutions aimed at claims and underwriting. Underwriting Insights and Claims Insights take the huge volume of data created by professionals in those positions, and use tools and techniques to turn it into actionable, forward-looking information.
For an underwriter, the program prioritizes, with machine learning, the most important applications in their queue to work on so top accounts that are the most likely to be valuable to the company will be worked on sooner. In claims, an adjuster can use the product to capture data from a claim and have it scored using predictive models on how complex the claim will be to resolve. Will a claim only require straight through processing or will it involve attorneys, injuries, or property damage?
“If so, let’s give it to our most experienced adjuster to handle this claim so we avoid a lawsuit,” said Rusch. “We’ll take in this claims information and we’ll score it across seven or eight different models, not only at the initial intake of the claim but also when any new information is added to the claim.”
Even though insurance is a business based on establishing relationships, there is pressure to adapt to a changing, and what some may call disruptive, digital environment. When the VP recently went to an insurer to implement the products, he witnessed some old fashioned ways of doing business.
“They are still using, believe it or not, paper reports and they’re still using Cobol Code and they’re still using antiquated techniques where someone actually goes in physically and runs a report, and then stores it on the hard drive for someone to pick up at a later date,” said Rusch.
The primary worry among insurers is that their affairs aren’t in the right order to implement AI or machine learning technology based on gathering data.
“What we hear a lot of is, ‘I hear you but my data is a mess,’ and we come in and say, ‘well you’re like most other carriers that we’ve talked to.’ Let’s help you organize your data and get it in a state where we can now start using it and ingesting it for building out machine learning and AI capabilities,” explained Rusch. “Yes, they’re a little fearful, but it’s more them thinking that their data isn’t in usable fashion, and we can help get them there so it is usable.
“You don’t need to have all of your data in alignment or all of your data in a data warehouse. We just need samples of data where we have known outcomes for fraud or known outcomes for underwriting experiences, and then we take those known outcomes and we can build models around those. Fear is probably more replaced by, how do I get started?”
