Now that virtual and photo-estimating have become commonplace in insurance claims processes, collision repairers likely are aware of the challenges it can potentially cause — undervalued vehicles, undervalued damage, extended length of keys to keys, et cetera. But have you ever thought about how artificial intelligence (AI) could help make fraudulent claims possible?

Using AI-generated photos of himself as examples in a recent social media post, Cambridge Mobile Telematics’ Ryan McMahon shared that the technology isn’t quite there yet to produce fraudulent photos in claims but it will be with time.

“Today, there are a number of applications that use photos to estimate damage from crashes,” McMahon wrote. “Generative AI applications like Midjourney can be used to enhance these photos in ways that Photoshop never could and they’re easy to use and only going to get better.

“Imagine a situation where small damage from a collision can be enhanced significantly to match exactly what these tools are expecting to see. In fact, it sounds [like] the perfect use case for generative AI. Now, all of a sudden the estimates far exceed reality.”

He also shared that he recently heard some great use cases for generative AI from Adrian Jones. Generative AI creates novel things such as artwork, music, and text which are not often used in insurance, according to Jones. He said algorithmic or predictive forms of AI are used in insurance to provide “straight-through” quotes and assist in claims processing. 

That got Repairer Driven News thinking, what could the ripple effects of AI in insurance claims handling be? So we spoke to Jones, who is an insurance investor, for some insight on the topic.

“It’s like virus/antivirus, meaning people use new technologies to create new frauds and insurers use new technologies to try to detect and prevent new frauds,” he said. “You’re already seeing it with AI, for example. There are sophisticated ways of editing photos using AI and there are ways of detecting edits.”

In addition, there are companies that say they can verify the time, date, and place of a photo or there’s always the option to have a video chat with customers to see the damage, Jones added.

“One of the questions that carriers are asking themselves is how much fraud, really, is there and what is the cost versus benefit of trying to stop it? A process might not be 100% perfect — the standard is ‘good enough,'” he said.

“If the savings on labor and the advantage to honest customers outweighs the potential for fraud with a new process or a new technology, then it might make sense, even knowing that some fraud might be introduced or allowed. It points to the old adage, ‘Know your customer.'”

A recently released eBook on AI processes in underwriting written by intellectAI, underscores that it’s “very critical to reimagine the future of the underwriting journey with a human-centered approach.”

“It is the human/AI partnerships that exhibit interdependence within the system that will be the most effective forms of partnerships in underwriting to produce better outcomes in terms of improved operation efficiency, effective better risk selection and enhanced customer experience,” intellectAI wrote. “…AI acts effectively as a triage tool, which can ignore information that isn’t relevant to the quote while only including data that is relevant, increasing the overall accuracy and quality of the work MGAs [managing general agents] are doing. MGAs have also continued to invest in data enrichment to be able to price and underwrite risks more accurately, which benefits everyone involved in the process.”

Jones told RDN AI regulations and standardization are needed to move forward in effectively using the technology in insurance.

“There’s a lot of really useful forms of AI throughout the insurance business,” he said. “However, many of them are not the sort of generative AI that has generated so much interest since ChatGPT came out.

“Insurers like standardization, and in many cases, regulation requires that insurers be completely standardized. The need for standardization limits generative AI in core insurance processes. There is not yet any consistent or widely agreed regulation at the state level or even at the NAIC around how to govern models to control the risk that they engage in unacceptable behavior. That’s a fundamental challenge for the industry.”

AI would have to be trained on very specific insurance information, which might be private or have restrictions around its use, to be standardized. That would also mean addressing varying existing requirements by state, Jones said.

For now, Jones said insurers have a few ways to detect and try to curb fraud:

• • Know the propensity of customers to engage in fraud — those with bundled auto and home insurance likely won’t risk losing their home insurance to engage in auto claim fraud; and
  • Consider raising claim deductibles. Jones mentioned home insurance carriers have done so to prevent soft fraud, such as in roof repair. For example, he said policyholders or roofing contractors claim questionable hail damage to get their roofs replaced.

CCC Intelligent Solutions Chairman and CEO Githesh Ramamurthy said last month that AI, specific to the collision repair industry through Estimate-STP, not only reduces cycle time but also administrative costs and environmental impact.

The application of advanced computer vision AI for claims processing increased 60% year-over-year, according to CCC’s insurer AI adoption report, released in February.

MarketWatch researchers found that the insurance industry’s use of AI may only increase due to customer demand and isn’t far from making real-time underwriting possible.

Seventy-nine percent of “tech-savvy” car insurance customers said they would trust a claims process fully automated by AI, according to a Solera Innovation survey, leading MarketWatch to believe insurance underwriting of the future may include a myriad of data points about a customer’s lifestyle and driving habits.

Images

Featured image credit: da-kuk/iStock