Can AI make debt collection smarter and easier?

Over the past decade, as banks have effectively outsourced debt collection to third parties, this corner of the financial world has slid into an abyss of harassing phone calls, shoddy recordkeeping and wrongful collections.

This is reflected in the latest data from the Consumer Financial Protection Bureau’s complaint database. Since the database was opened to the public in June 2012, Americans have aired more grievances about collections than about any other aspect of their financial lives. As of June, the CFPB had received 316,810 complaints about debt collection. The most common objection? “Continued attempts to collect debt not owed,” cited by 39% of filers. (The second-largest topic of complaints is mortgages, at 272,153.)

The CFPB has said it plans to hold banks and other first-party creditors responsible for accuracy of their consumer debt data even after it’s sold to third parties, which may accelerate the trend for lenders to bring debt collection back under their roof.

Banks historically sold distressed consumer debt to agencies, but that process was rife with problems. In August 2014, the Office of the Comptroller of the Currency issued guidance that dramatically reduced the pace of sell-offs. Since then, banks have primarily only outsourced the collection of the debt, versus selling it off. However, the CFPB’s actions could also curtail that practice.

Chart tracking complaints to the CFPB.

The timing is good, though. Artificial intelligence, chatbots and self-service technology have reached a point where they can provide a much-needed makeover to the collections process. Such technologies can help lenders learn to reach out to people at times and in channels that are more conducive to a conversation and repayment. They can help engage borrowers in a negotiation about their debt that can result in far higher repayments than repeated threatening phone calls.

AI at Acima
Acima, a provider of instant leases for online shoppers, drives collections with a homegrown AI engine, according to CEO Aaron Allred.

“The machine learning platform tells us who is past due, who’s most likely to pay, what’s the best time to call them, what time of day we should be calling them, or on what day of the week,” he said. “It’s the brains behind our collections efforts, so we’re calling the right people on the right day at the right time.”

Of course, the AI has to be trained.

“We have hundreds of thousands of customers we’ve made leases to, and probably half of them have missed payments somewhere along their lease cycle,” Allred said. “We start collection efforts on that. We dump that data into the machine learning platform.”

The system constantly analyzes what days past-due customers were called and how many times they were called. It also studies if they were called on their home, mobile or work phone.

“As you continue to dump more and more data into the platform, all those data points go back into the platform and it retrains itself,” Allred said. “It gets smarter and smarter as it gets more and more data.”

Debt negotiation portals
About five years ago, BBVA rolled out a collections self-service portal, where customers who are late on card or loan payments can go and ask for help, schedule a payment or make a promise to pay. In certain cases, borrowers could even renegotiate the terms of their debt.

BBVA did not allow an executive to comment on how well the portal is working today. Still, the idea made so much sense; we expected other banks to introduce similar sites, which seem like an obvious alternative or complement to collections calls.

While we haven’t seen others follow suit, they might start, said Steve Platt, president of Decision Analytics for Experian North America.

For starters, more banks are embracing the idea that they should nurture customers even if they fall behind in a payment.

“I think in the past, if a customer got too far behind on payments, that became a black mark that it would be hard to ever recover from,” Platt said. “With the sky-high costs of customer acquisition these days, managing and valuing your customers through that entire life cycle is worth more than ever.”

The wide-scale adoption of everything digital and mobile could also help.

“That means that not only would you use mobile for customer acquisition and managing your own banking relationship through an app, but equally so you can extend that technology into the collections space,” Platt said.

A third reason is that banks that have relied heavily on third-party debt collectors are starting to bring that activity back in-house.

Recently, Experian rolled out a portal similar to BBVA’s called eResolve that banks can license.

As a customer falls behind on debt, a bank using eResolve will send him a link to the service and give him a chance to work out the debt online, rather than through phone calls. The bank decides which options for bringing the balance current will be offered.

“It could be an option to simply resolve the debt there without having to have any interactions with collections group, it could be a payment plan the lender is willing to offer, it could be a reduction in finance fees,” Platt said. “It’s ultimately up to the lender.”

Embedded in eResolve and Experian’s broader debt management software, PowerCurve Collections, is a decision engine that helps the bank navigate choices about how to manage accounts that are past due.

It can drive decisions like how often to contact customers and the most effective way and times to reach them. Sometimes doing nothing is the right approach, Platt said.

TrueAccord’s virtual assistant
The fintech startup TrueAccord has built an AI-powered virtual assistant that several large financial services firms are piloting and might launch soon. It’s also working with LendUp, an online lender to people with poor or no credit histories.

TrueAccord creates a personalized strategy for each consumer.

“It’s proactively reaching out to the consumer in the best channel possible, it follows up, it reacts to consumer behavior, even if they look at the offer and drop off,” said founder and CEO Ohad Samet.

The virtual assistant is generally more likely to be empathetic than bullying, but it can be firm as well.

“People in debt are scared, they’re angry, but sometimes they need to be told, ‘Look, this is the debt and this is the situation, we need to solve this,’ ” Samet said. “Sometimes being too empathetic is not in the consumer’s best interest. People who are in debt can experience something that’s very neutral.”

Samet demurred as to what collections methods work best or what types of outreach dramatically improves the chances that someone will repay the debt.

“There isn’t one size fits all,” he said. “That’s why machine learning works where others don’t work.”

But he is a fan of a multipronged approach. There can a “halo effect” of reaching out to a consumer via multiple channels, he said.

“If you send a consumer two emails subsequently, you get one level of response,” Samet said. “If you send an email and then you make one phone call attempt — not five, just one — or you send one email and a text message afterward to follow up, the response rates can be as high as two to four times what the baseline of two emails will be.”

TrueAccord can also handle a negotiation process where somebody could change how often and how much they pay.

“If the consumer wants to pay $50 this week because they have two part-time jobs, this week they can pay $50, next week they can pay $25, the week after that they can pay $100,” Samet said. “That’s very convenient for consumers and something a call center can’t support. Customers want to feel like they’re heard and like they’re getting the customization they need.”

Typically TrueAccord collects 40% to 50% more than traditional collections agencies, Samet said.

The upper end of that range comes in the cases of prime consumers. “Some of them need reminders only, some of them are in financial duress that requires a lot of hand holding,” he said. “It really varies. But consistently we have a lot of data to show that we outperform traditional agencies.”

TrueAccord also has a collections agency subsidiary, but Samet stressed that agents there are not compensated for how much they collect.

“Our agents are paid to be accommodating, they’re trained to listen, they’re trained to work with the consumer, and the most important thing, they’ll get paid a reasonable salary no matter if the consumer pays or not,” Samet said. “They’re not on the phone saying, ‘Oh, crap, if I don’t make this I won’t get my bonus,’ which is the standard with a traditional collection agency. Which makes you wonder, why would banks submit their customers to this experience?”

Editor at Large Penny Crosman welcomes feedback at penny.crosman@sourcemedia.com.

For reprint and licensing requests for this article, click here.
Debt collection Delinquencies Artificial intelligence Virtual assistants Bank technology BankAI Conference
MORE FROM AMERICAN BANKER