Speed and data are insurers' keys to AI

To be ready for AI, insurers need speed, say executives for carriers and technology and AI providers to the insurance industry who spoke at InsurtechNY March 29 and 30. 

Speed, they say, applies to getting and processing data for insurance. Secondly, insurers need data capacity to feed AI for evaluating risk and making insurance decisions, according to the executives.

“Speed is one factor – if you can resolve a claim before there’s a legal process,” said Dan Dubiner, chief technology officer of Scalehub, a provider of automated crowdsourcing solutions. “Even with big data, one of the main challenges we have is the accuracy. How do you deal with all this big data to make sure that the prediction is very accurate?”

Counterpart, a management and professional liability insurer, partners with Markel and Aspen Insurance to back its policies – and also for data capacity, according to Tanner Hackett, CEO of Counterpart. 

“Speed and flexibility really matters,” he said. Counterpart’s partnerships “provide a lens into how technology can be applied to solve problems. Insurance companies are trying to accelerate their product development.”

Improving data processing speed, for instance, can set up creation of new insurance products, according to Jake Sloan, vice president of insurance at Appian, a cloud computing provider. “Being able to arbitrate sources of data rapidly without having to get in a spreadsheet – we see being able to create embedded products very quickly, align those to your core administration, and point to those rules if you want to augment the rules,” he said.

Robert Huntsman, chief data scientist, Prudential.

Prudential found that gathering internal and third-party data together could speed up decision-making for a new term life insurance product, according to Robert Huntsman, chief data scientist at the carrier. Having a better way to manage data helps comply with regulations for health data related to life insurance coverage, he added.

“There’s a lot of data we would like to use, which is very beneficial to the customer, like medical records, attending physician statements or electronic health records,” Huntsman said. “But we need to be very mindful of how regulators will look at how we use that information, even using the existing information from our application. We have to do significant testing to make sure that we’re not biased. It’s a combination of aligning our third-party data sources to allow us to first best execute the business strategy, but subject to the constraint that we have to also meet regulatory constraints.”

Sri Ramaswamy, founder and CEO of Charlee.ai

Pulling together more data quickly can also yield better risk insights, according to Sri Ramaswamy, founder and CEO of Charlee.ai, a predictive analytics engine.

“When you train AI over multiple different carriers, you are going to take those risk insights that have contributed to either the litigation or high severity or fraud,” she said. “These risk insights become very, very effective in predicting. When you do traditional machine learning, you can do that with structured data. That requires a lot of volume.”