Under the AI hammer?
When you have a hammer in your hand, everything can look like a nail, and there are few bigger hammers these days than artificial intelligence (AI).
AI is driving innovation — and concern — in sectors from colleges and universities to Hollywood, and lots of points in between. The insurance industry is no exception.
AI is identified, sometimes in the space of a single sentence, as both a transformative boon and an existential threat. So, which is it? And how can insurers and reinsurers resolve this and use AI to help chart a course through waters already roiled by dramatic changes in the insurance industry?
“Education is going to be critical if we’re going to use AI effectively,” said Grinnell Mutual President Dave Wingert. “There’s a lot of good the technology can do, but we also need to figure out how bad it could get.”
As one might expect with such a powerful technology, Grinnell Mutual has been deliberate about its exploration of AI. The attitude the company’s senior leadership team has adopted is one of optimism tempered by a cautious, let’s-see-the-numbers approach to change.
“This is gold-rush time for AI,” said Roby Shay, Grinnell Mutual’s chief information and chief operating officer. “We want to see if we can put some guardrails in before we start thinking about tying it into any of our core business processes. We think it would be wise to let things settle out and that we should align with major existing software platforms like Guidewire, where they’re already doing the heavy lifting in figuring AI out.”
CLEAR EXPECTATIONS
One way insurance companies can help themselves during this transitional period is to be very clear about what they want out of AI.
Doug McElhaney, a partner at McKinsey & Co. who focuses on AI’s effects on property-casualty and life insurance companies, said, “I think the thing an insurer really needs to keep in mind is the problem they’re most trying to solve for. If they’ve been struggling with how they’re adjudicating risk or handling their underwriting process, then focusing on leveraging more ‘traditional’ AI abilities like machine learning may provide the most benefit.
“But if what they want to address is customer service — streamlining their claims experience, for instance — then maybe generative AI would be the way they should go.”
A brave new world, right? But “new” can mean a number of things, some good, some maybe not so good. Innovative, groundbreaking. Untried, untested, error-prone. Or, all of those things.
THE GOOD
Used wisely, AI promises to be an important tool in the insurance industry’s effort to surmount the challenges that have beset it in recent years. According to a Carrier Management magazine article by Dennis Winkler, director of insurance for the Information Services Group (ISG), AI’s greatest potential will be the help it can offer insurers as they work to “unlock value and innovation, gain deeper insights into their clientele, enhance the precision of risk assessments, and elevate the quality of offerings.”
Paul Carroll, editor-in-chief of Insurance Thought Leader magazine, has for years kept a weather-eye on AI as it affects insurance. “Claims and renewals are where the rubber meets the road,” he said. “With claims, AI can gather all the info you need, ping the people you need to ping, and keep that process moving along. Before, it was a manual process that might take days, weeks, even months. It’s the same with underwriting — 500 possible data points, and the underwriter might look at 70–80. It’s hard to sort everything.
“With AI, though, you can grab all that info, prioritize it, and be more efficient. The biggest improvement so far has to do with renewals. AI can tell you what’s changed, but also put the things that are most important up front. The theory is that AI will enable better decision-making than humans can do on their own.”
Insurance software developer Guidewire has been using GenAI to help Grinnell Mutual streamline policy administration, claims management, and billing.
Amy Mollin, Guidewire’s vice president for product management, foresees insurers shifting from static workflows and forms-based systems to systems offering real-time guidance drawn from analyses of historical and current data. As befits the AI’s varied nature, she says, a combination of its techniques will enable this shift. Downstream, she said, GenAI will yield a vastly improved customer experience.
“It will enable new interaction patterns that are supported by the insights and recommendations machine learning provides,” she said. “We’re especially enthusiastic about opportunities in claims and underwriting.” Mollin predicts faster quote turnaround times, reduced wait-time in claims processing for both claimants and adjusters, and increased efficiency with straight-through processing.
THE NOT-SO-GOOD
For all the potential generative AI could have for the insurance industry, currently its promise is just that: a promise. According ISG’s Winkler, only 5 percent of AI-driven initiatives in insurance have yielded tangible value, and though insurance firms have received AI enthusiastically, most have confined their involvement to sandbox experimentation. Winkler said that 36 percent of insurance firms in Europe, the Middle East, and Africa are running “isolated projects,” while 25 percent claim they are working toward “transformation.” Only 2 percent have achieved full integration.
Why the slow pace of adoption? In a word, results. Or rather, a lack thereof.
“We mocked up some data that represented a problem and asked Chat GPT to give us some analysis,” Shay said. “In a couple of instances, what it confidently provided was just flat wrong. So, that experience gave us a perspective check. We know a lot of our people have been experimenting with Gemini, Chat GPT, and the free version of Co-pilot; so far, they haven’t showed us any big gains in productivity or business value out of that.”
This suggests that however “intelligent” AI might be, it will still be subject to that oldest of computer problems: garbage in, garbage out. Bad data or bad instructions produce bad results. This means human oversight of the tech is always going to be key if the industry is to stay on the right track with it.
“As our teams move forward with implementing AI tools, we need to ensure that humans remain in the loop, to catch any errors or biases,” said Nicole Chesmore, Grinnell Mutual’s assistant vice president for information technology. “AI can be useful in detecting fraud for example; that’s something it can do well. But human intervention will still be necessary to review the findings and make final decisions.”
For one thing, GenAI can be prone to hallucination. This is a phenomenon in which it sees patterns or objects in data that aren’t really significant, or are nonexistent, yielding outputs that are nonsensical or inaccurate.
Chesmore is confident the payoff for vigilance will be significant. Kept on a tight leash, AI “will give us opportunities to substantially improve customer service — for instance, using AI-powered systems that can provide quick responses, while maintaining a personal touch through human interactions when necessary.”
Wingert stresses that as the company moves down the road toward integrating AI into its critical systems cybersecurity will be a pressing concern.
“[With AI] there are scams and security risks that we didn’t have to face before,” Wingert said. “We have to stay a step ahead of cybercriminals, and to do that we need more sophisticated tools.”
BEYOND THE HYPE
“We don’t see [insurance] companies looking to replace people and jobs with generative AI,” said McKinsey’s McElhaney. “Rather, AI is going to enable companies to evolve their core processes, and enable carriers to look at how they’re going to make better decisions and unlock productivity.
“As generative AI continues to evolve, it has the potential to reshape roles and the associated tasks people in these roles execute. That’s going to allow companies to make better decisions and unlock capacity. That being said, it should also be said that AI is going to be disruptive.”
Grinnell Mutual’s Roby Shay agrees that a clear-eyed assessment of needs, goals, and expectations should be the starting point.
“While we’re careful not to get swept up by every new trend, we also recognize the importance of staying at the forefront of innovation,” Shay said. “Our goal is to adopt new technologies, like AI, when they’ve proven their value and can truly enhance our core processes. It’s about finding the right balance between caution and progress to ensure we continue delivering the best possible outcomes for our members and customers."
Shay references the Gartner Hype Cycle. “A technology is introduced, there’s a Peak of Inflated Expectations, then a Trough of Disillusionment, followed by the Slope of Enlightenment, leading finally to the Plateau of Productivity,” Shay said. “We’re going to wait until AI has reached its Plateau of Productivity before we implement.
“We’re people-centered and our success as a business is founded in our relationships. So, there’s going to be some skepticism for a while, and we’re going to wait until a lot of the dust settles before we make a judgement about what works for us, our members, and our customers.”
SIDEBAR
Despite AI’s ubiquity in current technology discussions, defining what “AI” means is harder than it might seem. In part, that’s because AI isn’t just one thing.
Artificial intelligence is an umbrella term that covers a wide range of technologies. Broadly speaking, it’s a branch of computer science focused on giving computers the ability to mimic human intelligence. At AI’s most basic level, AI-equipped computers can go where their programs tell them to, performing specific tasks by following specific rules. But they don’t create anything new.
Machine learning is another facet of AI in which statistical algorithms “learn” from data and extrapolate to perform tasks without explicit instructions.
Generative AI — GenAI for short — builds on machine learning models to simulate the creative process, producing new text, images, video, audio, or even software code in response to a prompt.
9/2024