Artificial intelligence (AI), which has made its mark in insurance in efforts to automate underwriting, rating, and claims adjusting, is now being deployed to read, categorize, and even modify the wording of insurance policy forms.

The potential impact of this development dominated a panel discussion of “AI and Insurance ROI” at the 2018 Global Insurance Symposium in Des Moines, Iowa.

Insurance carriers implementing any new application of AI can learn from recent experiences of their peers, said Ted Stuckey, head of the global innovation lab for QBE Insurance Group.

Too often, he said, companies develop pricing models created by actuaries and data scientists acting without sufficient attention to core business functions. “Very rarely do you see [the models] fully integrated into a start-to-finish work flow,” he said.

“I know of very few carriers who use analytics on the words they write. They care much more about the numbers.”

Ted Stuckey, QBE

“Artificial intelligence on its own is just a very complicated mathematical model that requires a high level of skill and sophistication to maintain,” he continued. “On its own it has zero value to carriers.

“The point of decision-making is not to have a bunch of actuaries and IT engineers sitting around a table. It involves underwriters. It involves sales people. It involves claims professionals. It involves product development people.”

Regarding product development, Mike Fitzgerald, a senior analyst with the technology research firm Celent, recalled how he wished one of his former companies had had AI resources for policy form analysis several years ago when it reviewed policy provisions throughout its portfolio.

“We had dozens of different clauses across our portfolio,” he said. “Some variations mattered, some didn’t.”

QBE had a similar experience more recently when it analyzed cyber provisions in its policies, said Stuckey. “We had well over 100 different cyber provisions,” he said. “Because there were so many, we couldn’t determine what our exposure was.”

QBE was hardly alone in that regard. “I know of very few carriers who use analytics on the words they write,” Stuckey said. “They care much more about the numbers.”

That’s changing. RiskGenius, an AI-powered platform that breaks insurance policies down into structured data, is one catalyst for change.

Among other things, use of Risk Genius eliminates the need to create “side-by-side” spreadsheets for policy form comparison. RiskGenius has also developed means for scoring different policies for comparative purposes.

Copying a clause and pasting it into a spreadsheet is not productive.

Chris Cheatham, RiskGenius

With AI-supported applications, “those people can spend a lot more time analyzing policy language, creating better language, and creating better products,” he added.

Stuckey agreed. By embracing artificial intelligence in product development, he said, “we’re going to create new, innovative products we can bring to market.”

Also, he added, with the enhanced efficiency, “we’re going to do a more thorough job of underwriting audits, and we’re going to work more closely with our underwriters and claim professionals.”

AI will provide both flexibility and structure to product management,” Stuckey added. “We will get to a point where an underwriter can write a policy to get a deal, but by using verbiage that’s approved in a manner that’s allowed.”

Stuckey acknowledged that won’t necessarily make conversations about utilizing AI any easier.

“You still have to ask people who are doing good work, hard work, and skilled work how long certain tasks take,” he said. “Then you tell them, ‘Well, this can do them instantly.'”

Nonetheless, he added, “the ROI conversation is one that has to happen.”