Credit unions captured growing segments of the auto lending market from banks
To help strengthen its portfolio, and reduce the stress on its workforce, Sierra Central Credit Union in Yuba City, California, enlisted the help of a Zest AI model to complement its longstanding partnership with Origence. The technology helped increase acceptance rates without raising delinquencies, while also freeing up underwriters to handle more complicated cases.
"The holy grail is you want to get more production without increasing staff because staffing is the biggest expense that we have, and I think that Zest plays into that very well," said Ernie Martin, senior vice president and chief lending officer for the $1.5 billion-asset Sierra Central.
Zest AI and Origence, a credit union service organization that specializes in connecting car dealerships to credit union financing, are adapting this technology for a white-label product called Zest Auto, which any credit union can use. The product, launched this month, combines Zest's underwriting models with Origence's customer origination platform.
The two fintechs are adjusting their focus to underscore the quality of decisions rendered by the algorithms, according to Mike de Vere, chief executive of Zest AI in Burbank, California.
"The issue of today's economy is that many credit unions are loaned out, so as we go into these uncertain financial times — whether it be a recession or not — the question is: How do we support a credit union and the dealer in making an accurate and smart decision?" de Vere said.
Zest AI honed the new product's efficiency by building a test model using consumer credit data from 2006 to run decisions on loans made between 2007 and 2008 during the Great Recession — eventually using the results to ensure fairness throughout all templates when reviewing applicants from underserved communities, de Vere said.
"We've got 250-plus models in production … so we need to take those learnings and make sure that we're applying [them] to modeling not just our current customers, but also our future customers," de Vere said.
Quarterly data from the National Credit Union Administration showed that outstanding auto loans, which include new and used, increased roughly 16.7% from 2021's total of $404.5 billion to more than $485 billion.
At Sierra Central Credit Union, new and used vehicle funding accounted for more than 56% of its $922 million lending activity last year. Martin stressed that more dynamic scoring is key for creating complete profiles for underserved consumers and better understanding an applicant's creditworthiness.
"The real power in the model is that it's able to identify those borrowers that are improving their credit. … So although their FICO score dampened down just because of what happened in the past, the Zest score takes into account" recent positive behavior from borrowers, Martin said.
But as helpful as automation is, analysts stress that proper oversight is crucial for navigating the
"An AI model's explainability is critical for regulatory compliance" and "regulators want to know why a model operates the way it does and why it makes an approval," said Craig Focardi, principal analyst for research and advisory firm Celent.
But regulators especially want to know why a model "either declines or recommends not to approve a loan," Focardi said.
Adopting tools for automation can require a certain level of trust from executives, said Daryl Jones, senior director at the Scottsdale, Arizona-based advisory firm Cornerstone Advisors.
There can be a disconnect when "the behavioral side never gets changed to adopt the technology and allow for the efficiency and scale," Jones said.
As
"Anyone who's been in lending knows it never feels good to say 'no,' and in the end, your hope is to [help] as many consumers achieve what they're trying to achieve in their life. … These tools will help credit unions do it more confidently," Hendricks said.
