Abound CEO Says AI-Driven Lending Still Needs People to Perform

Traditional consumer credit scoring, which overlooks a large proportion of the population, may have been the norm for years, but Gerald Chappell, CEO and co-founder at U.K.-based open banking startup Abound, says that needs to change.

“Traditional credit scoring is really good at separating your prime customers from your near-prime customers but works very poorly for those in the middle of the credit distribution,” Chappell told PYMNTS in an interview. “What that means is that millions of people, about a third of the U.K. population, get unfairly excluded from access to low-cost credit.”

To plug that hole, Abound is exploring a different underwriting approach, one that goes beyond traditional credit scoring and uses bank transaction data as the basis to better understand customers. “What that means is understanding our customers’ whole financial situation, what they earn, what they spend, and what’s left over at the end of the month,” he explained.

And that approach appears to be working, because in the last two years since the company’s launch in 2020, Chappell said their model has performed really well, producing a 70% reduction in defaults relative to the rest of the market.

“For every 10 defaults that our competitors are receiving in this segment, we’re receiving three, and that’s where we have that capacity to be able to offer our customers much lower interest rates,” he added.

Investors also seem to approve, and their recent injection of £500 million ($601 million) to help fund the open banking-powered lending business is a “good vote of confidence in the technology and the performance that we’ve had so far,” Chappell said.

The goal now is to be the “unsecured consumer lender in the U.K.,” he said, while leveraging the EU’s Second Payment Services Directive (PSD2) regulation, which provides access to similar open banking data as is available in the U.K., to lend to all consumer asset classes on the continent.

He said the British lender also has plans to expand into the business-to-business (B2B) lending space in Europe and is targeting between 100 and 200 business clients across the region within the next few years.

Extracting the Signal From the Noise

Abound’s approach to credit scoring uses a mix of open banking data and machine learning algorithms — a strategy Chappell said improves applicants’ credit scores and increases their chances of accessing cheaper credit.

He explained that historically banks have not been able to build accurate risk and lending profiles based solely on “messy” transaction data which involves “lots of transfers between multiple bank accounts and between peers and can be difficult to extract the signal from the noise in all of that data.”

Abound, on the other hand, leverages consumer transaction data gathered via open banking to build its artificial intelligence-based lending profiles, a process which he said enables them to “extract the complete financial picture from very messy and voluminous data.”

What makes that process even more unique is the fact that it uses a mix of machine (80%) and human intelligence (20%) to constantly refine the model decisions. Chappell said the human referral aspect allows the firm to make the edge cases that “aren’t super clean from the machine” smoother and consistently tap into the knowledge of skilled underwriters.

Another key element that drives a significant increase in predictive power is the ability to add new data into the engines using open banking, he said, pointing to how building a credit-decisioning model only using bank transaction data would probably get a slightly better, albeit equivalent level of performance to using just credit bureau information.

But combine that with a machine learning dataset and it’s a different ball game altogether: “That’s where you get a really significant uplift in power,” he said. “That uplift is what drives that 70% reduction in credit default rates I referenced earlier.”

Abandoning ‘Entrenched’ Underwriting Practices

While many traditional lenders and competitors are adopting the use of open banking-based data to smarter credit decisions, Chappell said it will take some time to get all lenders on board because of how difficult it is to do away with “entrenched” traditional credit scoring processes.

He predicts that it will take upwards of five years for a lot of mainstream lenders, including big banks, to fundamentally change their underwriting approaches.

In the meantime, however, Abound will be focused on perfecting its hybrid machine and human intelligence credit decision-making approach to get the right level of symbiosis.

“[That mix] gives a much better view of real-time affordability and allows you to make much fairer decisions. And in the end, fairer decisions result in better rates for customers,” Chappell said.

For all PYMNTS EMEA coverage, subscribe to the daily EMEA Newsletter.