Lauri Giesen
Lauri Giesen Apr 19, 2018

Artificial intelligence for real-world regulatory hurdles

Perhaps you’ll see this one coming like an angry regulator pacing up to the door. But if you wonder what federal mandates, a motorcycle, a baby boomer hairdresser and a blown business opportunity might have in common, read on.

Our story of regulatory woe on the go begins with 60-year-old Geri Michael of Vero Beach, Florida. The hairdresser decided it was high time to treat herself to a Honda motorcycle. Her checklist was looking pretty good, too. High credit score? Check. Successful business owner? Check. Unflagging loyalty? Check. (Michael had done business with the same large national bank for 13 years.) 

Aren’t banks always talking about how it’s all about the customer? Not this sterling customer, apparently.

Bankers told Michael that they’d had a bad history with motorcycle loans and so plunked her in a high-risk category—offering only to fund a small portion of her financing at more than 7 percent interest.

So off she went to Honda, and got her loan there, lickety-split, at under 3 percent with zero dollars down. And here’s how she celebrated: She rode back to her bank branch to show off her new set of wheels—offering her personal banker and other branch employees a spin around the block.

The same bankers who had declined her loan declined her ride.

Some variation of Geri’s story gets repeated every day in the financial services industry. And make no mistakes: Banks can’t fight regulators head on. They can’t bend the rules. But they don’t want to lose customers like Michael, either. That puts them in the unenviable position of working to please two masters—only one of which, we must note, makes the bank turn a profit.

“There has always been a trade-off between managing risk for banks and providing a good experience for the customers,” says Daniel Latimore, senior vice president of Celent.

In terms of data analytics, the scales have historically tipped towards regulation. Banks still use this tool to reduce bad debt in their lending operations while documenting how they comply with federal regulations.

But more recently they’ve been looking at new, more advanced analytics tools and technology—including artificial intelligence—to make sure they don’t just meet regulatory requirements but also gather more comprehensive information about loan applicants. Limited data can often cause risk-averse banks to automatically turn down loans out of fear. More data and better analysis helps them make better decisions.

“Banks are trying to get a better handle on their data,” says Robert Ashbaugh, executive risk management consultant for Sageworks. “They’ve always used data analytics to document requirements” that include CECL (the Current Expected Credit Loss standard) and Basel III (the international banking regulations designed to promote stability in the international financial system). 

“But now,” Ashbaugh notes, “banks try to understand the risk associated with loan applications to really quantify the risk they have. They’re also using it to see where their opportunities are.”

Many banks, he adds, don’t have as good of a handle on their loans as they think. But if they can reverse course, “not only will they limit bad loans, but also price the risk better. Good analytics allow them to slice and dice their data so they can ask, ‘Are we making money on certain types of loans and certain borrowers?’”

Artificial intelligence is already changing the equation in favor of banks.

“Banks are making better use of artificial intelligence to make smart lending decisions,” says BAI managing director Karl Dahlgren. “AI can be a win-win for banks and consumers. While it offers an opportunity for banks to grow and better price risk, the use of new models and data sources, if used effectively, could benefit under-served markets.”

That said, “The proper compliance protections need to be in place,” Dahlgren points out. “AI models constantly change and evolve, potentially in the wrong way, which could introduce disparate impact risks.”

Although AI is more commonly used in lending for fraud detection or customer service chatbots, some lenders are looking at forms of it to help them manage risk and predict customers’ ability to pay off loans.

“On the underwriting side, we’re starting to see some use of machine learning to help banks decide on loans,” Latimore says. “Advanced algorithms now take in all sorts of underwriting credit data and analyze customer behavior to make predictions about applicants’ ability to pay.”

But experts acknowledge the technology has a long way to go before it hits critical mass.

“I go to bank conferences and AI is one of the primary topics everyone talks about,” says Kevin Morrison, senior analyst for Aite Group.  “But in terms of implementation, most banks are directing their technology priorities elsewhere.”

The question is, why? Morrison agrees that non-bank lenders are more aggressive in using AI for credit risk management: “Banks are by their nature conservative and will wait to see what happens. But if they start to see some of the other lenders are finding success with this technology, we might start to see at least more testing of the technology by the larger banks.”

Besides using advanced technology, banks are strengthening their data analytics by broadening the range of information sources about applicants: a move meant to reduce bad loans and increase lending opportunities. “The more data you can gather about customers, the better,” Latimore says.

Morrison explains that banks have often relied too heavily on credit score and credit reports, ignoring other factors that would help them identify good loan candidates and exclude the bad. For example, many credit bureaus don’t factor in child or spousal support payments that could affect consumer debt ratios.

Yet banks can locate that information through other online sources. “There are many alternative consumer information sources out there and technology is makes it easier for banks to find those sources information,” he says.

There is even one source of data that is being talked about in lending circles that takes this concept a lot farther: use of social media.

“A lot of folks think social networking can provide a lot of information about customers.” Latimore says. “The thinking is that someone who has a lot of credit-worthy friends is more likely to be credit worthy.” Yet it’s still “really early” in terms of seeing how effective this source might be.

The bottom line—literally—is that banks, in finding better ways to evaluate loan applicants, should be able to manage their risk better and expand lending opportunities.

That way, people like Geri Michael can buy their motorcycles without getting mad at their banks … and ride off into the sunset smiling.

Lauri Giesen has spent more than 25 years writing about banking technology and payments for numerous business and financial publications.


Want more Banking Strategies? Sign up for our free newsletter!

For more articles like this, check out our recent Executive Report: "New trends in lending and mortgage."

With the right information delivered to the right people, you and your team members will learn more efficiently. That's why BAI compliance experts, including former examiners, have designed the BAI Mortgage Series specifically around the needs of lenders.

 

BAI Banking Strategies

Thank you for visiting BAI Banking Strategies. To view more, please Subscribe or Login.

Dismiss