The risks and rewards of full-scale AI exposure

IN Partnership with

Black swans and grey rhinos: how to prepare for the age of AI in insurance

IN TODAY'S increasingly complex claims market, artificial intelligence (AI) promises to be a potentially transformative force for companies worldwide. According to research from WorldMetrics, 67 percent of insurance companies have already implemented AI in their organizations, with 88 percent of insurers adding that they expect AI to improve the accuracy of underwriting decisions. But it is not all plain sailing. AI is still in its infancy – and leaders need to learn to walk before they can run. Raymond Ash, executive vice president, head of financial lines at Westfield Specialty, has a clear perspective on how industry leaders should approach the challenges posed by AI – particularly as states and countries seem poised to impose an increasing number of regulatory frameworks. “Everything begins and ends with the board’s responsibility of oversight of the company,” Ash tells Insurance Business.

Westfield Specialty launched in 2021 and, after its acquisition of Lloyd’s Syndicate 1200 in February 2023, Westfield Specialty closed year-end 2023 with $1.2 billion in GWP. Our experienced team brings deep expertise to the specialty market and offers unique insurance solutions for specialized risks. With nearly 400 employees across the US, UK, and Dubai, backed by Westfield’s over 175 years of financial stability, top-rated industry performance, and a well-capitalized balance sheet, we are positioned to be a leading specialty carrier.

Find out more

“It can be really challenging for executives to make sure that they stay on the straight and narrow with managing shareholder expectations”

Raymond Ash, Westfield Specialty

“Boards need to understand their own company’s potential AI exposures and ask smart questions to effectively identify and implement business opportunities while protecting consumers from potential hidden biases or unlawful discrimination.”

One of the most significant challenges facing companies today is how they communicate their AI strategies to shareholders and the public. Drawing a parallel to the early days of the COVID-19 pandemic, Ash warns that companies must be cautious not to overpromise or misrepresent what AI capabilities can deliver in the marketplace. “We’re in this same environment with AI,” he observes. The first AI-related class action lawsuit was filed in 2020; since then, the number of such lawsuits has continued to grow. “If you say you’re going to build this product or have a novel approach or build a better mousetrap and you don’t, it could lead to a potential lawsuit,” Ash explains.

This issue is especially relevant in the context of directors and officers (D&O) liability. The management team, tasked with designing and executing the company’s strategy, must also navigate the fine line between innovation and unrealistic or overinflated expectations. “It can be really challenging for executives to make sure that they stay on the straight and narrow with managing shareholder expectations,” Ash adds. Failure to do so can lead to potential securities class actions, where shareholders allege that the company misled them about its AI capabilities or potential.

Staying on 'the straight and narrow'

Published Oct 7, 2024

Specialty

Best in Insurance

Resources

Risk Management

News

RSS

Sitemap

About us

Conditions of Use

Terms & conditions

People

Specialty

Best in Insurance

Resources

Risk Management

News

RSS

Sitemap

About us

Conditions of Use

Terms & conditions

People

Specialty

Best in Insurance

Resources

Risk Management

News

RSS

Sitemap

About us

Conditions of Use

Terms & conditions

People

“An AI model is only as good as the team building it, the formulas and algorithms driving it, and the data it ingests. There is a whole concept of black swans and gray rhinos”

Raymond Ash, Westfield Specialty

Going further than internal concerns, the regulatory environment for AI is increasingly stringent, adding another layer of complexity. Ash points to recent regulatory developments, such as the SEC’s new rule requiring companies to disclose cyber breaches within four business days, as indicative of the broader trend toward stricter oversight. The question now turns to how materiality is defined, which varies across the regulatory spectrum and its potentially subjective determination. Best practices would suggest seeking the input of in-house and external counsel, the chief information security officer, and other information security professionals. Given these challenges, it’s clear that companies cannot afford to be complacent. Boards and management teams must be proactive in developing and obtaining the expertise and advice needed to navigate the AI landscape successfully. “[It’s about] ensuring [investment in] high-quality products which analyze data that does not contain hidden biases or lead to discriminatory results, making sure you’re really addressing the ultimate customer end use, and how that can go sideways,” says Ash. “We have seen what happened when technology mistakes lead to global outages that can result in dramatic repercussions. The same concept applies to AI if you have a novel approach to using the technology and the execution falls short.”

Increasing regulatory scrutiny of AI

AI’s impact on the insurance industry, particularly in the realm of risk assessment and modeling, is another area of significant interest. Insurers using AI for risk assessment have seen a notable 45 percent improvement in accuracy – reducing manual data entry errors by 85 percent. Ash touches on the potential for AI to revolutionize how insurance companies assess risk, particularly in property insurance. By leveraging AI to search for and summarize historical data, insurers can build more robust models and predictive capabilities. “I think that’s the most exciting opportunity right now,” Ash adds. “A lot of firms are looking at that, trying to do a better job of identifying and then building a better model. What they’re generally finding is that their models have to change

Identifying outdated technologies and improving best practice

so quickly in reaction to evolving global weather phenomenon where using AI technology can help them make better underwriting decisions.”

However, in the specialty casualty sector, which includes products like D&O liability, the human element remains central. In our everyday world, humans make decisions that can lead to mistakes with severe consequences. Sometimes, those consequences find their way to a courtroom, and human juries award nine-figure nuclear verdicts due to the emotional traits of a lawsuit. While AI technology is very strong at developing inferences from myriad sources of information, it has its limitations. And while the technology will evolve over time, “An AI model is only as good as the team building it, the formulas and algorithms driving it, and the data it ingests,” Ash warns.

Black swans and grey rhinos

“There is a whole concept of black swans and gray rhinos. There’s a risk in front of you that you’re not really paying attention to – black swans which are more aberrational and unlikely. But gray rhinos are something that are right in front of your face – something that can run you over if you don’t acknowledge and respect them. The people building AI models, through their inadvertent biases and/or the potential hidden biases in the data the models process, could unknowingly create black swans or gray rhinos if they are not careful.” In the insurance industry, claims management is another area where AI’s potential and limitations are evident. Advanced AI can help improve claims accuracy by up to 90 percent and even reduce errors by up to 25 percent. However, while AI can assist in summarizing data and predicting outcomes, the claims process remains fundamentally labor-intensive and reliant on human expertise.

And, as AI continues to reshape industries, the challenges for risk management and compliance are only set to grow. The responsibility for navigating these challenges lies squarely with the boards of directors and executive teams, who must be vigilant in their oversight and proactive in seeking the expertise needed to successfully identify and manage AI-related risks. The stakes are high, and the potential for missteps is significant. “Recently, enterprising plaintiffs brought a class action suit against a media company in California for using ads targeted to its customers, claiming customers were unlawfully discriminated against,” adds Ash. While AI technology wasn’t alleged to have been used in that case, “[It begs the question], if you are using AI technology to identify target customers, is this inherently discriminatory? To me that is the biggest exposure for companies, as well as if they don’t get the regulatory framework correct. “As much as we want to laud the concept and the technology, we’re a very long way from understanding how it’s going to affect customers.”

Understanding impact on customers

$1.3 trillion – value AI could contribute to the insurance industry through improved productivity and customer experience

up to 30% – reduction in operating costs of insurance companies possible with AI

95% – estimated customer interactions in the insurance industry that will be facilitated by AI by 2025

88% of insurers believe that AI will improve the accuracy and speed of underwriting decisions

How AI is helping streamline customer expectations

Source: WorldMetrics

up to 90% – improved claims accuracy as a result of advanced AI algorithms

up to 25% – reduction in claim errors due to AI-powered predictive analytics

15–20% – reduction in claims leakage AI can help insurers achieve

50% – reduced settlement times due to insurance firms using AI for claims processing

AI’s role in claims processing

Source: WorldMetrics