AI Is Forcing the Insurance Industry To Change During the Pandemic

Using data mining and prediction science to to better anticipate and assess risk. JL

Sarah Min reports in CIO magazine:

AI’s promise as a cost-saving, results-enhancing elixir might lift the insurance industry’s earnings. Increasingly, insurance carriers are turning to AI to assess risk and other costs associated with taking on a group of pension beneficiaries. The result may well be lower costs for companies that shift their defined benefit pension programs to insurers. Custom pricing, risk management, fraud detection and robotic process automation are sure to benefit by AI’s sifting through data.

Artificial intelligence (AI) is shaking up the insurance business, with implications for pension risk transfers, plus insurers’ stocks as an investment.

More than half a century since computer scientist Alan Turing proposed developing and testing intelligent machines, AI has made great strides. Processing costs have come down. Companies have access to bigger and bigger sets of consumer data.

The ability of computers to process, analyze, and learn from an elegant piece of code is disrupting all sectors of the economy, most recently in social media and retail, and also finance. Now, it’s insurance’s turn. That should be good news for the carriers themselves and their investors.

Pension risk transfers 

Companies offloading their pension obligations to insurers has been an escalating trend for years, one that only slowed recently owing to the coronavirus. According to the Life Insurance Marketing and Research Association (LIMRA) trade group, these buyouts dropped 46% in 2020’s second quarter, to $2.3 billion.

But that likely is a temporary situation. Once the pandemic is over, there likely will be a big snap back in pension risk transfers, says the organization’s Secure Retirement Institute. The ongoing corporate desire to lower retirement liabilities, plus higher expected DB plan payments to the Pension Benefit Guaranty Corporation, “will drive employers to seek PRT deals in the future,” said Mark Paracer, the institute’s assistant research director.

With the added PRT volume, the folks who engineer the transfers will need help. Increasingly, the insurance carriers that provide PRTs are turning to AI to assess risk and other costs associated with taking on a group of pension beneficiaries. The result may well be lower costs for companies that shift their defined benefit pension programs to insurers.

After the United Kingdom changed the tax rules in 2015 to increase peoples’ access to their pensions, consultancy Mercer started considering a new tool to help consultants choose options benefits transfers they find most appealing. 

Since Mercer started testing its pension risk transfer AI tool, which it launched in January, it found that current pensioners and future beneficiaries respond to different packages. Both care foremost about the size of the package. The algorithm helps calculate a size that is both attractive to beneficiaries and cost-efficient for pension funds. 

The consultancy also found that the not-yet pension recipients are most likely to accept a package in the spring. Analysis shows that the timing of the offer is important, especially for members that have not yet retired. The lesson is to “send offers to members at a time that we know they are more likely to engage – when members engage, they then have an opportunity to make an informed decision about their pension income,” said Maurice Speer, principal in the risk transfer group at Mercer.

“By its very nature, the more exercises you do, the more it learns, the more powerful the predictive ability will be going forward from this,” said Andrew Ward, head of risk transfer at Mercer.

Insurance Stock Investors: Boost Ahead?

Over the past five years, stock prices for life and health insurers have dropped, thanks to investors who are concerned about low returns.  Lincoln National’s stock price has tumbled 32%, MetLife’s has dropped 9%, and Prudential has fallen 14%. Lower rates and COVID-19 claims are expected to add greater pressure to the sector. 

Meanwhile, investors have long been bullish on property and casualty (P-C) insurers, which started the year with record levels of capitalization and have recovered since the market plunge of the first quarter. Since October 2015, stock price for Allstate has more than doubled at 56%; Progressive has climbed nearly 197%, while Travelers has jumped 8%. 

But the stock market’s applause for P-C comes with an asterisk: Continued uncertainty from the coronavirus and historically low interest rates promise trouble ahead. So P-C insurers will be under mounting pressure to reduce claims costs where they can, such as by modernizing their claims processing. If they can’t come though, look out below, investors.

Which is why AI’s promise as a cost-saving, results-enhancing elixir might well allay any concerns, and lift the insurance industry’s earnings, helping both woebegone life companies and popular, if fragile, P-C outfits.

AI Aims to Sharpen Insurers’ Profitability

Insurers, who live and die by actuarial tables, damage probabilities, and other statistics, are sure to benefit—as are their bottom lines—by AI’s sifting through data. Here are other ways AI’s is disrupting the staid insurance industry, in a bid to make carriers more efficient: 

Custom Pricing. So long as data privacy regulations don’t get in the way, custom pricing is in the future of insurance companies. Algorithms analyzing custom data will price risk more accurately to distinguish between two customers who might otherwise be the same: they could share the same age, gender, and neighborhood, but have vastly different risk appetites that can impact their auto, health, or travel insurance. 

A crop of pay-per-mile auto insurers in recent years are implementing machine learning into their processes by incentivizing drivers to fit gadgets into their cars. How often they drive, how many miles they log, how often they brake, and how quickly they accelerate can provide a customized premium that may be attractive to customers. 

Risk Management. Meanwhile, in health insurance, AI is trying to predict mortality and illness prevalence to determine the size of claims or the seasonality in claims. Many factors go into who falls sick, how bad and expensive it will be, and who may die.

Fraud detection. Machine learning will also increasingly help insurers curb bottom line losses to fraudulent claims. That blueprint is already used in other businesses. One global credit card company found that stolen cards are often swiped at fuel stations, prompting the firm to zero in on them. They can target the problem more precisely and send text messages to unsuspecting card holders to alert them to fraudulent activity, according to Aanand Venkatramanan, head of ETF Investment Strategies at Legal and General Investment Management. 

But not all fraudulent transactions are so simple. For example, someone who accidentally damages their car and parks on the side of the road may make a claim in the morning, which insurers may not be able to confirm is false or genuine. That’s where huge data sets reviewed by AI come in. What are the chances this claim is bogus? What is the average claim size? How many similar claims have been filed over the years? 

“I would not have a personal history of insurance claims or you might not have yourself, but people like us would have had claims in the last 50 years,” Venkatramanan said. “You can pore through that data to then say what is the probability of me making a claim being genuine or fraudulent.” 

So long as the false positives are fewer than the true negatives, huge savings could be brought to insurance companies, he said. 

Robotics Process Automation (RPA). In the future, claims can be automatically processed, saving humans the drudgery and time of reviewing them. For both claimants and insurance companies, it could result in a better user experience and cost savings. Firms such as Automation Anywhere and Blue Prism offer these AI services. 

Data Privacy Concerns 

The downside, of course, is that this torrent of AI-driven data could end up hurting  consumers and insurance firms. In North Dakota, Insurance Commissioner Jon Godfread, who chairs the AI working group at the National Association of Insurance Commissioners (NAIC), said the fund is working on guiding principles for the emerging industry. They are following cues from the finance sector and the global insurer industry to curb risk from machine learning. 

There are no standards for algorithmic auditing yet, according to Godfread. That could expose companies, who implement untested machine learning across their books of business, open to risk. Like algorithms that inadvertently discriminate people by race, ethnicity, and gender, such a weakness could open insurance companies to class action lawsuits and other financial risks.

Example: An insurer that offers better premiums to families living closer to attorneys’ offices and daycare centers may economically discriminate against minorities. 

“We’re doing our very best to make sure we’re rooting that out of any system that’s coming forward,” Godfread said. 

Due diligence is necessary. Third-party auditors exist for companies to review algorithms to ensure they don’t unfairly shut out claimants, but the process likely needs further regulation. More clarity should emerge in the next year, as insurers, regulators, and other wrestle with these ethical questions. 

For now, global insurers and investors will do well to review AI thoughtfully and carefully.