Can AI help banks thwart elder abuse?
Banks are stepping up their efforts to detect and deter financial elder abuse in response to a rise in such crime, and artificial intelligence software could become part of the solution.
“Wells Fargo has been focusing on this as an issue and building analytics for it, very much like we do for other things like fraud,” said Rich Baich, Wells' chief information security officer. “We’re greatly concerned, and we’re putting time and resources behind it,” including having teams of data scientists create proprietary models.
In 2018, U.S. banks reported 24,454 suspected cases of financial elder abuse, a 12% increase over 2017, according to the Financial Crimes Enforcement Network, which is a unit of the Treasury Department.
These numbers could be an undercount. Banks do not have to report exploitation of the elderly on suspicious activity reports unless they meet a certain threshold.
“The problem is, no one is mandated to do it, and if they’re not mandated to do it, no one is going to do it because there could be negative ramifications,” said Larry Santucci, senior research fellow at the Federal Reserve Bank of Philadelphia. “For example, you don’t want to be the only one reporting on your elder exploitation cases. Banks and financial advisers are in a difficult position. Even the ones that want to help can’t muster the fortitude to share this kind of data and report it.”
Banks should be allowed to share this kind of information with each other, he said.
Cases 'every week'
Bank executives and industry observers say that not only is this type of fraud under-reported, but it is also growing rapidly.
Laurel Sykes, senior vice president and chief risk officer of Montecito Bank & Trust, started to notice an increase in elder scams in 2015 and built a department to deal with them and related crimes.
“We used to see a handful of cases a year — now we see a handful of cases every week,” Sykes said. The majority of the bank’s customers are baby boomers and older.
“We see the grandparent scams, and we’re seeing more of the romance scam,” Sykes said. “Our older customers are either divorced or widowed, and they’re out on dating websites being scammed by people pretending to be someone else.”
The bank also sees work-at-home scams, where customers who think they are hired for a job receive a check that turns out to be bogus.
Moreover, it sees grandsons, daughters and other family members who are supposed to be taking care of the older person actually steal from them.
In one case this month, the bank suspected a daughter-in-law of taking a client’s money. The bank filed its suspicions with adult protective services.
“The son- and daughter-in-law kidnapped our client, took her out of the state and ran off with her money,” Sykes said. “But because we’d done our filing, law enforcement was able to step in, and they got caught.”
Research shows that elder scams are usually perpetrated by loved ones and caregivers, according to Liz Loewy, former chief of the elder abuse unit in the Manhattan District Attorney's Office and current co-founder and chief operating officer of EverSafe, a provider of software for detecting senior scams.
Judy Long, president and chief operating officer of First Citizens National Bank in Tennessee, said the bank filed more SARs for elder financial exploitation in 2018 than ever before.
Some of her customers have been told they have won a lottery ticket and asked to provide their account information to receive the payment. Others have been scammed by family members.
Santucci pointed out that the over-65 population in the U.S. grows by 10,000 every day, so even if only 3% are affected by elder financial exploitation, the numbers will continue to grow.
Loewy noted that older people hold close to 80% of the assets in the U.S.
Tech approaches today
In addition to educating branch staff and customers about the signs of elder abuse, many banks use existing fraud detection, anti-money-laundering or Bank Secrecy Act compliance software to find them.
Such programs are typically rules based. A rule might be, if the number of debit card transactions is greater than three and they happened between midnight and 3 a.m. and the person is over 70 years old, flag that account for potential fraud.
Montecito Bank uses the BSA software from Verafin to analyze the accounts of clients over 65. It tracks if an account balance is decreasing, and if that jibes with past activity. It analyzes account access to see if new individuals have been granted access to the customer’s account. It looks for transaction types uncommon to that customer, such as suddenly wiring money out of the account.
The bank’s information technology department is also working on a way to detect scams in which seniors are talked into buying gift cards and phoning in the numbers on the back.
Last year, Long at First Citizens National Bank combined the bank’s fraud and BSA departments and deployed a transaction-monitoring system from Bankers Toolbox (the company recently changed its name to Abrigo) to monitor accounts of customers over 65, looking for patterns of fraudulent activity.
"We know our monitoring tools are catching that early so we can set up visits and conversations with those customers to find out how it started, why it started, and hopefully begin to prevent that,” she said.
Need for AI
Rules-based transaction monitoring software can see what happened, but cannot predict what is likely to happen.
“The big problem with transaction monitoring is it typically relies on a big drop in balances before the alerts go off,” Sykes said. “We’d love to get out in front of it before the balances have dropped.”
What is needed, Santucci argues, are programs that can compute the probability that a person will be involved in a fraudulent scam or financially exploited.
He cites EverSafe as one example of AI-based software that can detect "diminished financial capacity, a loss of executive function that prevents you from performing your day-to-day banking tasks.”
Before an older person becomes susceptible to elder fraud, such as a grandparent scam, their banking behavior often becomes erratic, he said.
“Your banking transactions go all over the place,” Santucci said. “You’re forgetting to pay bills, you’re forgetting your password, you’re making mistakes in your banking that you then cover up by bringing in money from other accounts.”
Banks can analyze such behavior to build predictive models to determine the likelihood that an older customer will be defrauded or financially exploited.
“You need to do predictive analysis to save people money before the money goes out the door,” Santucci said. “The technology is there to get in front of it.”
EverSafe uses predictive models to build a profile of each senior banking customer’s behavior. Then the software looks for deviations from that profile.
The software sends alerts out to the senior and trusted advocates — family members, financial advisers, lawyer and accountant. An alert might flag the absence of a pension check that is normally deposited at a certain time each month, for instance.
“That transparency can be a deterrent in and of itself,” said Howard Tischler, co-founder and CEO of EverSafe.
Tischler started EverSafe after his mother was scammed. A telemarketer sold her an auto club policy when she did not have a car or a driver’s license and she was legally blind. Then other telemarketers sold her inappropriate things, and she stopped paying her long-term-care insurance, withdrew money early from her annuities, and had a friend who was helping her write bills who would write checks to herself.
"She lost her lifetime of savings as a result,” he said.
One advantage EverSafe has over banks’ internal efforts, according to Tischler, is that it can analyze transaction behavior across accounts at multiple banks and investment advisers.
“The fact of the matter is that this stuff happens across accounts,” he said.
Fidelity and Raymond James are both on-the-record EverSafe clients; some banks use it but will not talk about it publicly, Tischler said.
Wells Fargo uses its own machine-learning models.
“We have a team of dedicated data scientists who build model-based, machine-learning capabilities to identify anomalous and suspicious patterns to protect our clients,” said Ron Long, director of elder client initiatives for Wells Fargo Advisors. “We continue to refine our approach over time by adding new sources of information, both structured and unstructured data, to strengthen detection and prediction of potential concerns.”
Machine learning is required to analyze the hundreds of thousands of transactions Wells Fargo processes daily and find the 400 or 4,000 that require a closer look, he said.
"While a tool can’t replace human assessment, machine-learning capabilities play an important part in our strategy to reduce the number of matters requiring a closer look so we can focus on actual cases of financial abuse," Ron Long said. "This is what the industry needs and what we are continuously striving towards.”
Judy Long said First Citizens National Bank recently began experimenting with AI software that mines debit card and online banking data to detect fraud.
“We would like even better software to help us solve these cases and pinpoint elder fraud even faster,” she said. “On digital channels, fraud can happen so quickly. If fraud patterns could be detected quicker, we would like that.”
Editor at Large Penny Crosman welcomes feedback at email@example.com.