Insurers to tap technology for future successReprints
Technology is slowly changing the way insurers do business, but they have to adapt more quickly — or risk being swallowed whole by the machine, as Charlie Chaplin was in the classic comedy “Modern Times.”
“The insurance industry is probably the last major industry to adopt some of the newer technology and I think part of the reason for that is that it’s always backwards-looking,” said Gloria Vogel, an adjunct professor of finance at New York University and an insurance industry analyst.
“You look at the claims and then you get an idea on how to price the risk.” The rise of the machines was recently highlighted with Brian Duperreault’s appointment as the new CEO of New York-based American International Group Inc. and the announcement that Attune, an online insurance platform, will expand to target a market segment of companies with annual revenues up to $35 million, representing up to $150 billion in annual gross written premiums.
Attune’s creation was announced last September by AIG, Pembroke, Bermuda-based Hamilton Insurance Group Ltd. — where Mr. Duperreault had been the chief executive prior to returning to AIG — and Two Sigma Insurance Quantified, an affiliate of Two Sigma L.P., the hedge fund firm affiliated with Hamilton.
Hamilton, which Mr. Duperreault joined in 2013, expanded significantly after he took over the leadership of the firm. Its U.S. platform makes extensive use of technology in underwriting, including a business owners policy that relies on data rather than extensive underwriting questions to offer coverage quotes.
Francois Ramette, a Chicago-based partner with Strategy&, a PricewaterhouseCoopers L.L.P. company, said Attune may become “more aggressive, more ambitious in terms of the goals that it has established for its platform, but the underlying trend has been there for quite some time.”
“It’s always a question of, yes, you can achieve this, but if you can only do this for one out of 1,000 submissions that you get, you’re not changing the economics of the industry,” Mr. Ramette said. “But then again if they manage down the road to do this for a very large percentage of the risks or submissions that come to them, then it would be a pretty big achievement.”
Seth Chandler, Foundation Professor of Law at the University of Houston, said that “by taking large amounts of data into account, computer-devised algorithms can often be more accurate, predicting risk better than human rules of thumb or history, or older methods.”
“So there’s a lot to be said for use of machine learning and things like neural networks in underwriting,” Mr. Chandler said. “There are downsides, and the downsides include that usually these algorithms are inscrutable. There is no human on earth who can understand what the machine is actually doing; all we can say is that it seems to be predicting well, but if you try to explain what it is doing, it’s basically impossible.”
However, even predicting can become, well, unpredictable.
“There is also an element of arbitrariness in these algorithms — that is, you can often have several algorithms that are equally accurate in aggregate,” Mr. Chandler said. “They’ll get it right 90% of the time. But in individual cases, particularly the hard cases, the machines will give different answers. One machine might say you’re high-risk and the other machine might say you’re low-risk.”
But insurance underwriters don’t have to worry about being replaced by the machines — at least not yet.
“I would say the only way that could happen is our world remains static, but that’s not our world right now,” said David Kodama, Chicago-based assistant vice president of research and policy analysis with the Property Casualty Insurers Association of America. “We’re constantly seeing new risks and exposures, and it’s challenging the insurance industry in many aspects and many ways. And we’re going to see that challenge increase, and that’s why we need these tools to help our workforce to be able to address the new exposures.”
Greg Williams, senior director with ratings agency A.M. Best Co. Inc. in Oldwick, New Jersey, said underwriters will be needed for complex risk.
“I think right now, especially on the commercial side, they have so much data they’re almost overwhelmed by it,” he said.
“So I really think the next stage of (technology and artificial intelligence) is giving the underwriters a better way of interpreting some of the data that they’re getting, streamlining the reports so they become more useful for the underwriter.” Mr. Williams added that the role of the underwriter will likely change to more of a sales role, “trying to identify potential clients, only in a marketing way, armed with the information that they have.”
PwC’s Mr. Ramette said there will always be human underwriters to at least do a quick check to make sure the algorithms have worked properly.
“At least for now, there’s going to be a pilot in an airplane,” Mr. Ramette said.
“Even if that person is just going to make sure the landing is happening appropriately, there’s always somebody there. I think we can make the same analogy for commercial underwriting.”