Shortly after leaving the San Diego Sheriff’s Department over a decade ago, Pat Beaty became a certified fraud investigator.
“Fifteen to 20 years ago, large banks, financial institutions, insurers, and credit unions would have 15-20 people analyzing raw data,” he says. Since then, a lot has changed. “We’re still making calls, interacting with people, verifying members and their transactions by telephone. But looking at a lot more data.”
Yet despite the exponential increase in information for his company’s 80,000 members, Beaty’s company stays on top of fraud thanks to software using machine learning, which scans giant volumes of data to isolate suspicious activity.
See how Intel is using artificial intelligence to identify more fraud rings than ever before, saving companies money in the process.
“Fraudsters’ and money launderers’ use of advanced techniques, including AI, is accelerating. Cutting-edge AI isn’t optional—it’s required for investigators to be effective in anticipating and responding to these changes.”
Right now, many criminals act with impunity. The United Nations claims that less than 1 percent of global illicit financial flows are frozen or seized, and that up to 5 percent of global GDP--$5 trillion annually--are money laundering transactions.2
“Fraud is in so many different arenas—it impacts every industry as well as human traffic, money laundering and the funding of illicit activities, organized crime, and terrorist financing,” said Don Fancher, U.S. and global leader of Deloitte Forensic.
The insurance industry is a particularly lucrative target, as it collects almost $5 trillion in premiums each year.3 Fraudulent claims account for $80-100 billion annually in the U.S. alone.4
In the world of auto insurance, new nefarious practices are emerging all the time. Criminal activity ranges from lone actors to sophisticated fraud rings, sometimes involving staged accidents and medical providers in on the scam. The traditional methods for identifying suspicious information have failed; one leading insurer avoids paying out only 0.33 percent of the estimated 10 percent of fraudulent claims.5
This insurer deployed Intel AI for a pilot program, having it study one year of claims data in one state. Using its “reason by similarity analysis,” which looks for similarities and patterns in claims, the AI was able to illuminate patterns that were previously invisible to investigators.
Out of 113,000 claims, it was able to identify one ring involving 38 claims, $400,000 in fraudulent billing, and 42 participants, including medical providers such as physicians, acupuncturists, physical therapists, and psychologists. Within 10 hours of setup, it found another ring was bilking the insurer’s membership out of $2 million a year.6
The potential cost recoveries are enormous; the insurer estimates that for every 0.1 percent decrease in claim payouts, they save $10 million.7
There are several reasons why the AI outperforms other investigative methods. It’s not just superior processing power; traditionally, many leads go cold because of legal constraints or overburdened human resources.
“Investigations can be hard because of the jurisdictional boundaries for a lot of local law enforcement, and federal law enforcement is already backed up against the wall with their caseload,” said Beaty.
That means that even if sleuths had the bandwidth to parse troves of data to identify patterns, they lacked the access to information to do so. Intel AI is able to use anonymized data to overcome these traditional boundaries.
“Fraud is in so many different arenas—it impacts every industry as well as human traffic, money laundering and the funding of illicit activities, organized crime, and terrorist financing.”
“We have found fraud rings that had been resident in the data for months, if not years. That’s because unifying multiple heterogeneous sources and doing rapid similarity analysis over the information was not previously possible,” said Gayle Sheppard, vice president and general manager for Intel® Saffron™ artificial intelligence.
She explains that this new type of AI is available to address evolving criminal tactics.
This is welcome news to the insurance industry. A recent study by the Coalition Against Insurance Fraud (CAIF) found that a majority of insurers believe there has been an uptick in fraudulent activity in recent years.8
“Fraudsters’ and money launderers’ use of advanced techniques, including AI, is accelerating. Cutting-edge AI isn’t optional—it’s required for investigators to be effective in anticipating and responding to these changes,” said Sheppard.
CAIF spokesperson James Quiggle says that insurers understand the answer lies with technology.
“Crunching big data is key to the future for insurers. The United States is a giant black box--we’re a hyperconnected society that’s churning out discoverable data in vast quantities,” he said. “We’re entering a golden era of analytics. Artificial intelligence is evening the playing field against increasingly complex fraud rings, which are invading the landscape in ever-growing numbers.”