Adoption of AI is highest among top-tier carriers, according to a LexisNexis survey of 300 executives at the top 100 U.S. insurers.
The company’s “Hype or Reality: The State of Artificial Intelligence and Machine Learning in the Insurance Industry” white paper reports that more than 80% of respondents from the top 20 insurance companies meet the definition of “adopter” for AI and machine learning. The percentage dips to 62% for ranks 21 to 50 and 51% for the rest.
By line of business, auto insurance is leveraging AI and machine learning at a 68% clip, LexisNexis found, followed by life, commercial and home insurance. Despite the incursion of many AI-focused insurtechs, 65% of companies with active implementations, pilots or approved projects prefer to develop their applications internally. Marketing, underwriting and claims are the most common practice areas in which AI is being applied.
John Beal, SVP of data science for LexisNexis Risk Solutions, says that the most common implementation of AI at insurers tends to be computer-vision initiatives like image recognition for claims.
“The ability to read an image is becoming very common,” Beal says. “We don’t see the image recognition piece as the key, the challenge is pulling that data in from the image as a set of attributes, then combine it with a whole bunch of other attributes like property characteristics.”
There are four major challenges insurers face in adopting AI, Beal adds: Financial, staffing, data and regulation. The first two, however, are reported as much bigger pain points than the latter as carriers are generally still getting up to speed with AI.
“When we look back five to 10 years ago, trying to hire data scientists, insurers used to compete against each other, but now it’s every other kind company that wants to do analytics of some type,” he explains. “The other challenge is that once people come in, their expectations are so high.
“The regulatory piece is ramping up more in the next few years,” he continues, noting that officials will be looking at “bias evaluation, privacy, consent, data governance” as they evaluate insurers’ AI efforts.
