Alternative Finances

Why AI Risk Tech And Banks Could Be A Match Made In Heaven


While most CEO stories are about having founded a firm to solve a problem, Upstart CEO Dave Girouard and his team of former Google staffers actually started with a solution. A solution, he told PYMNTS four years ago, that had nothing to do with financial services or lending at all — they were actually studying hiring and figuring out ways to track employability over time with data.

But it occurred to them that their solution was useful outside of HR — and that many of the things that made someone a good hire of over time could also make them a good credit risk over time, if the artificial intelligence (AI) model they were using to screen with were modified to that task.

And “luckily” for Upstart, there was a big problem in desperate need of such a solution: the millions of thin credit or bad credit consumers who were locked out of mainstream systems. And since access to credit is a need to have for consumers, not a nice to have, it seemed a problem well worth taking on.

“Our simple idea was that better technology and better data science could improve access to credit,” Girouard said.

Seven years down the line, he said, Upstart is profitable, reports defaults at half the rates of competitors that lend to similar customers when rated by FICO score and has just raised $50 million in new funding to expand beyond its core personal loan business.

Expanding Access to Credit With AI

Except for a fairly small minority of consumers, everyone more or less relies on access to credit. The FICO-dominated system that is mostly used to evaluate credit worthiness, however, is a problem because it just doesn’t capture consumers' worthiness to borrowers as well as it could, Girouard said.

Ability to repay, he noted, is a fantastic criteria by which to assess whether or not to give someone a loan — and the credit industry has a long and ignoble history of using people’s inability to handle credit as a tool for harvesting fees and interest revenue. But, he said, the standards that FICO sets miss large sets of modern  consumers who could make excellent borrowers — but they are typically locked out.

“Young people, self-employed people, recent immigrants — the tools available often end up being very blunt measures to answer a very sophisticated question about the risk associated with making that loan,” he said.

What industry a person works in, how long they’ve been continuously employed, their salary growth patterns and how much education they have are all valuable data points. When they are filtered through smart AI that is learning all the time, along with “fancy math,” it allows Upstart’s system to approve more borrowers at a lower cost than traditional systems can because the Upstart system is more accurate in assigning risk.

And, as is traditional with online lenders, Upstart can do that online, in a couple of minutes, after a consumer has filled out an application. Over the term of the loan, he observed, the firm tends to see its customers’ traditional credit scores increase by 20 to 30 points on average.

“We only do personal loans, as of yet we don’t do mortgage or auto loans and we are just testing out a credit card product with one of our banking partners. But in lieu of that we are helping customer to be able to get whatever the next product they need is at a better price,” Girouard said.

Which, he noted, is not to say they are averse to someday, hopefully in the not too distant future, offering their customers those products themselves.

Expanding Offerings and Partners

Today, Upstart is a profitable firm, which means it didn’t need to raise $50 million to fund operations. But, Girouard said, after proving its underwriting with personal loans, the firm is now reading to expand — and wants to hire the data scientists and engineers necessary to build out those other underwriting types. The time is coming, he said, to get beyond the personal loan and into the other flavors of credit that are critical to consumers.

Upstart is also looking to continue to expand its roster of banking partners — which as of its latest funding announcement has already been growing rapidly. Customers Bank and its BankMobile division, First National Bank of Omaha, First Federal Bank of Kansas City and Accion Chicago are all slated to partner with Upstart to extend a white label version of its lending platform to their customers.

Banks, Girouard said, are natural partners for Upstart, because as competitors they will always carry a massive advantage to the race that he said any startup would be hard-pressed to overcome — low cost of capital.

“Over the long haul banks will dominate consumer lending, and we think our technology combined with the funding base for a bank is the best of both worlds for the banks and the customers,” he said.

Because, he said, banks may be the force that dominates consumer lending, but there are no guarantees the ones that dominate today will still rule tomorrow. Innovation, he said, is still an advantage, and Upstart’s goal is to bring its accuracy in judging risk to the advantage of as many banks as possible going forward.



The How We Shop Report, a PYMNTS collaboration with PayPal, aims to understand how consumers of all ages and incomes are shifting to shopping and paying online in the midst of the COVID-19 pandemic. Our research builds on a series of studies conducted since March, surveying more than 16,000 consumers on how their shopping habits and payments preferences are changing as the crisis continues. This report focuses on our latest survey of 2,163 respondents and examines how their increased appetite for online commerce and digital touchless methods, such as QR codes, contactless cards and digital wallets, is poised to shape the post-pandemic economy.