Just as SAT scores don’t convey everything about a student, FICO scores don’t convey everything about a loan applicant. They can be one valuable piece of information, but when banks rely solely on scores from the credit bureau to decide which businesses or individuals can be trusted with a loan, they omit much more information than they include — perhaps missing the opportunity to build a relationship with some of the most creditworthy applicants out there.
That’s why Dr. Akli Adjaoute, CEO, Brighterion, said that far more data must be brought into the decision process. It’s not enough just to prevent people from using stolen or synthetic identities to get a loan, he said; financial institutions (FIs) must discern the true trustworthiness of legitimate customers, even if they have never done business with those customers before.
Adjaoute said doing this goes far beyond what FIs can offer through manual, paper-based underwriting processes, which can keep applicants waiting for a month or longer. This, said Adjaoute, is a job for artificial intelligence (AI).
He said AI can give FIs a fuller picture of loan applicants faster than traditional methods. In a recent episode of the PYMNTS “AI Myths” podcast series, which explores credit risk and default prediction, Adjaoute told Karen Webster how true AI, applied correctly, can boost core functions of a financial institution when it comes to loans and more.
When Less Isn’t More
Evaluating credit risk isn’t just about keeping fraudsters from getting loans under illegitimate identities. Even when dealing with legitimate customers, said Adjaoute, a lot of information factors into someone’s ability to pay back a loan — FICO credit score, yes, but FIs can also learn a lot from demographics, transactions, debt-to-income ratio and even social media.
Adjaoute said AI is a next-generation technique for assessing the risk posed by a consumer or merchant, as it can combine all data from across sources to get a total picture of someone’s creditworthiness in real time — which, of course, is what applicants have come to expect.
For business loans, Adjaoute said AI can analyze context such as region and demographics to determine whether a certain type of business can succeed in the location where it’s trying to set up shop.
“If you really have AI,” he said, “it’s more than statistics, neural networks or data mining; real AI deals with personalization and is adaptive and self-learning. You should be able to analyze news, data feeds and any other type of information in addition to data from the credit bureau.”
Serving the Unserved, Yet Creditworthy, Customer
Speed is important, said Adjaoute, but the real value of AI is the ability to identify high-quality applicants who may not look high-quality based on their credit score alone.
For example, Adjaoute remembers moving to the U.S. in the early 2000s. He had a successful company in Europe at the time, but he had no credit history in America, struggling even to get approved for a phone. Nobody knew who he was.
Now, 15 years later, his daughter is emerging from a master’s program at UCLA — but because she wasn’t working during school, her score with the credit bureau may appear less than impressive.
Adjaoute said that’s not just frustrating because she’s his daughter; someone who has been in school will probably soon find a job, and someone with a master’s degree is even likelier to find herself in a leadership or management role.
Those are exactly the types of applicants FIs should want to approve, Adjaoute said — even if the customer’s credit looks “bad” today by a FICO score measure; the odds of it improving over time (and thus, of the applicant eventually repaying the loan) are extremely high.
Furthermore, he said, AI allows FIs to get much more granular with credit risk and default prediction, enabling them to distinguish between customers who will simply default on their loans versus ones who may be delinquent for a while but will ultimately pay up.
Adjaoute added that Brighterion’s system is able to reduce defaults by 76 percent by predicting the riskiest applicants versus the ones in the middle who, despite sometimes struggling to keep up with payments, still represent profit and a positive relationship for the bank.
Of course, choosing who receives a loan is only half the battle; FIs must also keep a finger on the pulse of that loan, because things can change. A business that started out strong can run into financial troubles, while one that struggled in the early days can get its feet beneath it and thrive.
With AI, the same intelligence that informed the decision to issue the loan can keep the FI up to date on positive or negative developments to help avoid any unpleasant surprises down the line, said Adjaoute.
“It’s easy to lend money,” Adjaoute concluded. “The trick to successful financing is getting it back and making more money off what you provided. That’s why it’s time to advance the existing solution to bring in the new types of data that are available.”