RCBC (Rizal Commercial Banking Corporation) in the Philippines is now piloting the artificial-intelligence services of a Singapore-based fintech, Bizbaz, to predict borrower behavior, says Lito Villanueva, RCBC’s chief innovations and inclusion officer.
The central bank, Bangko Sentral ng Pilipinas, has given commercial banks until the end of this year to convert 50 percent of transactions to digital and enroll 70 percent of adults into the formal financial system.
That’s a big goal considering about 44 percent of the Philippines’ adult population (of a total of 117 million people) is unbanked, according to McKinsey.
RCBC is a universal bank founded in 1960 as part of the Yuchengco family conglomerate, with a large business serving micro-, small- and medium-sized businesses. In 2020 it launched a digital platform for rural customers. But many potential customers are still out of reach.
“Nine out of ten SMEs are micro-SMEs, with no profile, so they lack access to credit,” Villanueva said.
RCBC didn’t apply for a fully digital banking license, and Villanueva says RCBC wants to leverage its branches as well as provide a digital service. So it is using fintechs like Bizbaz to extend its reach in the digital arena.
Piloting Bizbaz
“We’re not using Bizbaz to replace our credit process, but to enhance it,” he said, noting tech can cut through the administration and deliver credit decisions close to real time.
The bank selected Bizbaz for a pilot because it’s backed by HSBC and is working with other banks in Southeast Asia. RCBC is testing it against benchmarks such as the quality of new accounts it generates, time to market, the speed of processing credit decisions, its reliability, and how much RCBC can save in costs.
But what is going on beneath the hood of Bizbaz’s AI? What exactly does it predict?
According to the company, it predicts someone’s willingness to repay a loan.
Measuring character
Credit officers throughout history have sought better ways to make such predictions. As John Pierpont Morgan said in 1912, “The first thing [in credit] is character.”
Because humans and our systems can’t really tell what’s inside someone else’s head, banks are conservative, often refusing to lend to anyone except the customers who don’t need the money.
That might be all right in a developed market but it’s a barrier in countries with stubbornly high levels of unbanked people.
Has fintech cracked this challenge? Old banker Morgan would have relied on corporate financial statements as well as his exclusive network to make a lending decision. Credit bureaus made this available to all banks, but they require borrowers operate within the formal banking sector.
Fintech has since augmented this, mainly by generating huge amounts of alternative sources of data about a borrower. Such scoring techniques either provided different assessments of someone’s economic or financial activities, opening a window into their situation. Or they relied on their behavior: if they paid back a small starter loan, then give them a bigger one.
These devices are still ultimately about using tech to get a better look at past behavior or current status. But what about predictions, not just about someone’s ability to pay, but their willingness?
Predicting behavior
This predictive element is not just about technology. It’s rooted in a field of behavioral finance called psychometrics, which is the art or science of measuring someone’s personality traits. Combined with artificial intelligence, it becomes a tool to predict a person’s willingness to pay.
These tools build on longstanding categorizations of personality types. Early versions of this required borrowers to fill out a lengthy questionnaire to determine their personality traits. The advent of the smartphone enabled fintechs to harvest measurements of personality at scale – such as from a phone’s metadata.
Such data requires customer consent (although such niceties vary across markets), so psychometrics is also a driver of consent-management protocols that are becoming necessary to open-banking models.
Although these scores are becoming more accurate, they raise a number of issues. The biggest one, though, is that a score is only relevant so long as the customer continues to use that same device.
In many emerging markets, however, people and business owners tend to change phones or SIM cards regularly. New phone or chip, and the data history is a blank.
Banks using these fintechs haven’t been able to build reliable credit histories.
“The portion of Southeast Asians who are underbanked hasn’t changed in ten years,” said Bizbaz co-founder Hayk Hakobyan. “Fintech has had no impact on financial inclusion because it’s based on the phone.”
The other challenge is that banks that do receive valuable alternative data may struggle to integrate it into their credit-decisioning, risk and compliance systems.
The ultimate source
Bizbaz uses voice technology to try to get around this problem. Its voice-recognition software enables it to build a credit profile of a person with just thirty seconds of their speaking into a phone. It doesn’t matter what they say, according to Hakobyan. While his company combines this with other data types, sometimes – for poor villages in remote locations – the only way to obtain data is from voice.
“There’s no alternative for the poorest people,” said Hakobyan. If a country wants to make financial inclusion a reality for its entire population, he says, its banks must use solutions such as this.
He is a proponent of incorporating different measurements into a lending decision, not just financial. For example, in emerging markets, a high proportion of adults suffer from diabetes. Disease correlates to defaults. Bizbaz tries to combine its insight into financial behavior with such factors to help it work out who’s willing and able to pay back a loan.
For more sophisticated and affluent customers, a multiplicity of such factors combine with the psychometric aspects to provide an accurate predictor. Is someone conscientious? Are they impulsive, a chancer? Hakobyan says the tool is used by human resources departments to gauge C-suite candidates, for example.
But at its most basic level, when such data is unavailable, the AI’s voice diagnostic can be enough for simple lending products, such as a buy-now, pay-later service. It wouldn’t be enough for big-ticket products such as a mortgage or a life insurance policy.
Pscyhometric pitfalls
Psychometrics has its detractors. The first fintech to deploy it was LenddoEFL in the US, but it eventually stopped operating there and now focuses on emerging markets. Why? Because it is not clear what is a predictor versus what is just about someone’s innate character. Moreover, it was easy to allow gender or racial attributes muddy the picture, even if unintentional. This led to charges of bias and discrimination.
Other issues involve privacy and safety. Psychometrics is meant to require customer consent. But do people understand what kind of data they are sharing, or the way that insights are generated about themselves?
Hakobyun acknowledges the legitimacy of these issues, but says they are being met in two ways.
One is through tech. Bizbaz builds in ways to calculate personality traits based on duration. For example, the AI can pick up if someone is suffering from burnout. This could make them a risky customer. But such emotions are short-lived, so the AI calculates this factor may only last a few months. But an adult who scores high or low for conscientiousness is probably hardwired that way, at least for 10 years or more, so this type of factor will weigh more heavily.
The other is regulation, which goes back to fintech’s role in driving open banking. Some Asian governments are adopting aspects of the European Union’s General Data Protection Regulation (GDPR), which is designed to enforce data privacy and protect consumers.
Some factors may not change quickly in Asia. For example, GDPR forbids asking someone’s gender, but this is a valuable component in calculating personalities and will likely continue in Asia. But other aspects of GDPR are coming into play, as regulators work to ensure people trust open-banking regimes.
“I expect open banking to become the norm,” Hakobyan said. “We will see a greater focus on consumer consent management. This is the way we make financial inclusion a reality.”
Financial institutions with an inclusion mandate are keen to see these models take off – whatever the technology enables. RCBC’s Villaneuva said, “We don’t want to just know about AI, but live and breathe AI, because it’s the next big thing.”
