Upstart, an AI-based lending marketplace provider that traditionally served consumers who couldn't get credit elsewhere, is shifting its attention to the kinds of prime borrowers that banks embrace. On Thursday, it's launching a program called T-Prime that will let bank and credit union partners target super prime borrowers, using Upstart's marketing channels and automated lending system.
For more than a decade, the San Mateo-based company has used artificial intelligence to analyze consumers with low or no credit scores, for instance recent college graduates with little credit history, and tried to identify "future prime" or "hidden prime" applicants who are actually good credit risks by looking at other data elements including the types of jobs they have and their salaries.
As Upstart has forged lending partnerships with more banks and credit unions, it's become clear that these financial institutions are most interested in making prime loans to the most desirable borrowers.
A case in point is Alliant Credit Union, a Chicago-based, digital-only credit union that serves employees of large companies and association members across the country. The credit union, which has no branches, has been working with Upstart for about a year, offering debt consolidation loans to consumers who have racked up a lot of credit card debt.
"The aim Upstart now has to serve some of these more prime tiers is a strong overlap with our interest in serving and acquiring new members," said Dennis Devine, president and chief executive officer of Alliant, in an interview. "We generate new members, we generate deposits, but we also want to be effective at extending high quality loans to our members. That's where a partner like Upstart can fit. And if they are focused on serving members that we're able to serve and extend credit to, the partnership becomes a pretty interesting one to us."
Upstart is now consciously targeting prime borrowers. It's also adjusted the price ranges the model can handle. It recently launched a new algorithm, Model 18, that incorporates annual percentage rates of loans.
"APR is normally part of the output of a model: What APR should I charge this customer based on all these factors we know?" said Upstart CEO Dave Girouard in a recent interview. "But the interesting thing is that the APR itself affects the performance of the loan."
If a customer is charged a higher APR, their monthly payment is going to go up and therefore their likelihood of default also rises.
But there's also adverse selection at play: A person likely to accept a 15% loan is typically less creditworthy than someone who's only willing to accept an 8% loan. By taking APR into account in the credit decision, Model 18 has been boosting credit performance, Girouard said.
Devine declined to share how many loans the credit union has made through Upstart's platform.
"We started small," he said. "We wanted to see that it was going to work. We wanted to see the member experience and we want to see what the credit performance looks like. They've been quite good."
The benefit of using an AI-based loan decision model, rather than a more traditional FICO-score-based underwriting system, is "the more elements of data that you have around an individual allows you to make a much more thoughtful decision around the risk that that individual might present for the likelihood to repay or not repay," Devine said. Upstart also adds economic forecast data that factors into a borrower's ability to repay, he said.
Alliant monitors the model's activity for risk, compliance and performance, he said.
According to Upstart, the key benefit a platform provider like Upstart brings to banks and credit unions is a fully automated process.
"The process of originating an online loan, especially if you are a new customer who doesn't already have a decades-long checking relationship with the bank, is generally very difficult," said Paul Gu, chief technology officer at Upstart, in an interview. "It often involves going into a branch. It most commonly involves uploading documents. In our case, 90% of loans are instantly verified."
And working with Upstart can help banks acquire new customers around the country, he said.
"That's been a pain point we've heard from banks and credit unions: How do I acquire net new customers in a digital native way?" Gu said. "That is not necessarily something that comes easily. And it's something that we've spent the past 10 years refining, an ability to do it cost effectively."
Banks decide their pricing structure, in other words, target APRs, as well as the risk range levels they are willing to accept.
"Through those parameters that they set, we have all of the information that is necessary from a pure mathematical standpoint to then determine APRs for any given applicant," Gu said.
Upstart itself is still interested in subprime borrowers, Gu said. But it's acknowledging that prime customers are a huge part of the addressable market.
Though AI-based lending has existed for more than a decade, it has yet to take off among mainstream banks and credit unions.
A big factor in this hesitancy is regulatory uncertainty, according to Christine Livingston, managing director and global AI leader at Protiviti.
"Underwriting credit risk decisions is subject to a lot of specific regulations," Livingston said in an interview. "Regulation is probably one of the primary reasons more organizations and banks haven't used that type of platform." Also, for many banks it's hard to change loan underwriting systems that are typically a component of a complex core system.
"You've got to migrate legacy data and restructure and reorganize," she said.
But there's also reason to think demand for AI-based lending will increase, she said.
"The best experience anyone has anywhere becomes the minimal expectation for the experience they want everywhere," she said. "I can go on Amazon and I can get literally anything I want delivered to my door, probably within 24 hours. And if I have an issue with that, I can really easily fix it. I can return it. It's an incredible experience."
Consumers will start to expect this kind of ease from their banks, she said.
"The customer proposition of, oh, I don't need to sit down and sign my life away on a hundred different documents and wait for a week to see if my loan is approved, I definitely think there's a really interesting consumer angle to that," Livingston said. "I can see that being very appealing from a consumer perspective and for banks that are looking for efficiencies in the underwriting process."
