As artificial intelligence budges its way into the business world, insurance companies are grappling with the changes to the kind of work that happens, the types of workers they need and how to develop them. This part 2 of WGLT’s series on AI and the insurance sector. You can also read Part 1.
AI will affect several broad areas of the insurance sector, fraud detection, risk assessment and damage assessment. Gunratan Lonare, a professor at the Katie School of Insurance and Risk Management at Illinois State University, said in the future, AI will take over low-risk claims assessment.
“Instead of a human looking at this small stuff, AI can see, OK, what's the length and size and age of the house? It can capture all the data. It can also capture the context. Was the roof damaged because someone hit a stone, or because of a hailstorm or something else?” said Lonare.
AI can also assess the severity of damage, he said. This reserves extra human capacity for more complex damage cases.
Fraud
Insurance companies have been doing data analysis to root out fraudulent claims for a long time. Pete Miller is the CEO of the Institutes Risk and Insurance Knowledge Group, a not-for-profit organization which educates and connects people in the industry. Miller said AI can improve that analysis to root out organized crooks in a big way.
“There are sophisticated fraud rings in several parts of the insurance value chain around perhaps attorneys or doctors,” said Miller. “There's a patterning that everything's going to this doctor and this attorney ring.”
Increasingly, some foreign governments in eastern Europe and other places are getting into insurance fraud as part of an overall suite of cybercrime initiatives.
“There are some nation states that are supporting hackers that might set up fraudulent websites,” said Miller. “They can make search engine optimization such that it goes to this fraudulent website, and the person actually thinks it’s their insurance company.”
The customer enters their data on the fraudulent site, and it gets sold on the dark web. He said the hackers typically set up these websites in advance of a severe weather event like a hurricane. The scary part, he said, is that's happening even before an insurance company gets notice of a loss or claim. That makes it harder for insurance companies to detect, although at least one AI company has a tool that may help under that scenario, he said.
ISU’s Gunratan Lonare said “good AI” is competing with “bad AI” in a kind of arms race.
“People can use AI to come up with exaggerated accidents, even fake video…fake images. If you have small damage, AI can make it bigger damage, and AI can completely change the video,” said Lonare. “Now, the question is, how your AI is better than the fraudsters' AI?”
Risk detection
Insurers have for some years had the ability to monitor driving habits of policy holders with sensors the policy holders may agree to have on their vehicles. AI will make those “telematics” of how safely and how much people drive more robust, according to Miller. AI will also help insurance agents develop better risk profiles for home policy holders. That can produce variations in rates. Miller said it can also be used to prevent losses.
“I've seen simulations around wildfire, for example. …Here's your house as it exists today, including the surrounding vegetation. Here's the impact of wildfire coming through. Conversely, if you remove vegetation for a safety barrier, here's the impact on your house, and it's likely to be better,” said Miller.
Competition for workers and talent development
The insurance industry has a long-running workforce shortage and challenges in attracting new workers to a business perhaps unfairly stereotyped as boring. AI could both help and hurt that situation. Lonare said AI technology shifts may increase the competition for some of the more marketable skill sets. For instance, data scientists and IT professionals in one field can move to another with relative ease.
Julia Lamm is a principle in workforce solutions for PricewaterhouseCoopers International who specializes in the insurance sector. Lamm said insurance companies are risk averse and may not be adopting AI as fast as businesses in other sectors. Lamm said the early adopters will have an advantage in hiring the kind of workers needed in this new AI world. On the flip side, since AI does away with many routine repetitive tasks, Miller with The Institutes said insurance may escape its stereotype as an unexciting field.
The change in the nature of work to do this fraud detection, create effective software and check on AI processes poses new questions for insurance companies about how they develop employee expertise, the fund of experience on which an employee bases judgment calls and transfers knowledge. For example, if you don’t come up by serving your time in the claims division for a while before moving on to another area, companies will have to find another way to teach workers about their unique business process.
Lamm said no client of hers is saying they want to step away from the apprenticeship model. She said she believes insurers will develop talent in a more accelerated way from the bottom, perhaps assisted by AI.
“Rather than learn on the job being your primary and, in some cases, almost sole way to learn, they're thinking about a lot more creative learning paths, simulations, sharing lessons learned in a different way. Now you can create trainings very, very quickly,” said Lamm.
Lamm said it’s easy to push those modules to people who need them. This minimizes the cost of sharing knowledge. She said most clients believe workers will move from entry level to productivity faster because information is democratized.
“If you have someone who's getting a new challenge, maybe it's an underwriter who's underwriting a type of risk they've never underwritten before. Instead of just going to the specialist above them, they can do very rapid learning on their own to understand the baseline and then call on the specialist…you know, check my assessment, help me think about what I missed,” said Lamm.
Before, the person who's more experienced at the top was telling them everything, said Lamm.
Miller said companies will have to be very intentional about information transfer because there is so much of it.
“Do you have data fluency? Can you understand and look at that data and sort of say, here's signal, here's noise? You have AI literacy. Do you know what that model does?” said Miller. “And just a mindset around sort of looking at things holistically.”
He said work will require a critical thinking, problem-solving approach. That worker quality has always been treasured by employers. Miller said now it’s crucial.
“There's the old adage of, ‘start at the end and work backwards,’…that I think is going to be foundational because the tools in education in general are going to be so sophisticated,” said Miller.
At least right now, colleges and universities are still teaching mostly in what Lonare termed a “traditional way,” offering subjects like insurance, coding and finance in isolation. He said it is “concerning.”
“That's where the education system needs to update their coursework. OK, we will be teaching you the basics, but at the same time, we will be teaching you how to use these AI tools to kind of make it faster, make it more efficient,” said Lonare.
Lonare said he is developing what may be the first such AI-centered course in the nation [AI applications in finance, risk management and insurance] and it’s perhaps a year away from rollout.
Middle management
AI also has implications for management, particularly middle managers. Lamm thinks there may be a need for a bigger middle-level of generalists who are doing a broader range of oversight than in the past.
Middle managers have never had it easy. They have managed people. Now, they will have to manage processes and exceptions to processes, said Miller.
“Management is going to change quite a bit. It really collapses layers in an organization, right?” said Miller. “The stuff in the middle, it's good, it's accurate, it's very fast. Most of that stuff goes through, but the manager has to use discretion and judgment on the exceptions.”
That doesn’t mean people skills won’t still be at a premium. Miller said there’s a continued need for coaching because critical thinking is a skill that takes time to develop.
Lamm agreed. She said understanding people plus understanding where “AI work slop” has gotten through, are tall tasks.
“I think the role ends up becoming harder in some cases,” said Lamm. “In many cases, we're starting to see models where there's a manager over more employees than they've had in the past, and so because that employee is now being helped more by AI agents, I think the role of the middle manager remains quite challenging in the future.”
Country Financial said it remains focused on helping employees build new skills through ongoing learning, upskilling and reskilling opportunities.
“The company continues to invest in future talent through its internship program, which this year includes 68 interns including 38 Country interns, 11 Illinois Farm Bureau interns and 19 University of Illinois DigitaLab interns."
State Farm did not offer a detailed comment on this point.
Who wins
With any huge technology shift there will be winners and losers. Steam engines and the industrial revolution jolted civilization in the early to mid-1800s. Around 1900, electricity and the internal combustion engine came along and changed global society in an even bigger way. New businesses grew. Old ones fell. Miller said AI is different from those disruptions because it’s the first tool designed “to create or replace human cognition.”
Even the internet didn’t do that. He said one commonality with earlier waves of change is that the people who have the technology and are able to leverage it have a significant advantage.
“To the extent that that causes consolidation, I think that that is that is very significant. And in an industry that's highly regulated, like insurance, that's a significant barrier to entry. I think there'll be consolidation around that,” said Miller.
He said the massive amount of data needed to effectively predict and prevent insurance losses, predict new risks and accurately price policies will be a significant advantage. Big insurance companies may get bigger and tend to crowd out midsize and small carriers.
The perception of unfulfilled promise will also inflect the business competition question. Lamm with PwC said right now is the middle of a hype cycle. First came excitement. Implementation began. Now, doubts are creeping in because of uneven results caused by imperfect or incomplete adoption of the tech.
“I've got clients who will say anecdotally to me now, ‘I started to use AI for a certain process, and when we start looking at all-in costs, it's not just the cost of the tokens; it's the cost of the governance to manage that agent. It's the cost if a new regulation comes through. How do we now have to update the agent accordingly?'” said Lamm.
She said the clients that respond to those costs and pressures thoughtfully will reap larger gains later. Right now, though, she said a number of insurance companies are leaving a lot on the table.
“If you're thinking just about use cases, you're not thinking about how processes change. You're not thinking about how rules change. You're not thinking about the change management and how people are going to come up this learning curve. Those companies are being heavily underwhelmed by their results,” said Lamm. “They're missing the full investment and the technology is amazing."
She said AI technology is moving fast. It's the people who can't keep up. And the companies that are more successful are prioritizing bringing their workers along in learning how it all works.
