Artificial intelligence is helping reduce bias from dealers and lenders who might otherwise underestimate or overestimate a borrower’s income in the auto finance process. Reducing that bias speeds up deals, lowers risk and matches borrowers with the right type of loans, said Jessica Gonzalez, director of auto lending strategy and account management for Informed.IQ.
AI, widely used by lenders making upfront credit decisions for car buyers, is a relatively new tool for income verification, Gonzalez said. The company, based in Tiburon, Calif., works with seven of the top 10 auto lenders, including Ally and Capital One.
Instead of using AI, some lenders still “stare and compare” income documents to confirm the information dealers submit to them. Gonzalez said using AI to verify income especially helps car buyers who are nervous about walking into a dealership, such as lower-income and thin-file consumers with little credit history .
“That kind of nervousness is really escalated because they’re walking in not knowing if they’re going to get funded, or is this dealer going to take advantage of me, or am I going to pay for a lot of extra things that I don’t need to get done,” Gonzalez said.
Bias in the F&I process can result in customers receiving add-on products they didn’t want or need, not knowing their true credit score or being charged more than necessary for an interest rate.
Estimating a car buyer’s income is not always easy, especially when the buyer is a gig worker who might not be including commission in their total pay. In addition, lenders each have different policies on how to calculate income. Discrepancies can occur if one lender looks at two pay stubs vs. three, or if a lender looks at pay stubs from a 60-day time frame vs. 30 days. Calculating income using bank statements, especially when data is input incorrectly from the start, creates different outcomes.
“Whenever you have AI and you see so many different data points, so many different variations of a W-2 pay stub — you become more effective on how to calculate that income with the minimal amount of risk,” Gonzalez said. “You open up the opportunity [for funding] to a market that maybe historically had underrepresented their income.”
Low-income consumers are most at risk for bias. Informed.IQ data shows auto loan borrowers from the low-income segment are underrepresented about 43 percent of the time.
Informed.IQ has portals where dealers and consumers upload documents and receive real-time feedback on how income is calculated to provide transparency. This process tells customers which deals and vehicles fit their qualifications, Gonzalez said.
Bias in calculating income for car buyers can result in troubling situations.
If a buyer’s income is overrepresented to a lender by just $5,000, that can impact the interest rate and terms. Once the income error is discovered, the consumer would no longer qualify for the loan and the dealer might have to claw it back. If the customer drove the car off the lot and the dealer can’t find another lender to fund the loan, they have to tell the customer to consider another car.
“When you walk out of the dealership, you think your loan’s closed, you’ve done all the good things, you’ve signed the paperwork. But in reality that loan is not funded,” Gonzalez said. F&I managers “would have to go and search for another lending organization to provide this loan. So they’re calling all their friends in the credit unions or anywhere they can and saying, ‘Hey, what can you do for me?’ ”
Car buyers’ increased demand for transparency, along with the COVID-19 pandemic that pushed online car shopping, sped up the need for upfront proof of income. Trusting the technology, understanding its terminology and ensuring regulatory compliance are still challenges associated with AI.
Using AI doesn’t require training for F&I staff. The main benefit for lenders, who do receive training on how to use the AI tool, is easy-to-understand income data.
Lenders judging dealerships on the performance of the loans they deliver isn’t always fair because a dealership can’t control their customers’ loan payment success. But if a small dealership in a low-income ZIP code with only a few deals consistently gives lenders clean deals that don’t cost a lot of money to onboard, lenders are more apt to do business with them.
Craig Courtney, finance director for Taylor Chrysler-Dodge-Jeep-Ram in Bourbonnais, Ill., said using AI for income verification is timely.
“It will be a good tool for lenders at a time when they’re trying to automate everything as much as possible,” Courtney said. “If it is accurate and it will help with the lenders’ loss ratios by making sure they’re doing a good job of verifying income upfront, that’s great.”
For example, AI verification would confirm reported income when a customer can’t immediately show a pay stub, Courtney said. But he also pointed out potential challenges.
“I can also see hiccups if AI is off and didn’t calculate properly and now a person [with an 800 credit score] is getting ‘stipped’ for income and gets irritated with us for even asking,” he said, referring to added requests for stipulations. “Overall, it will be fine, but it will be an adjustment for the consumer if there’s either bias or miscalculation.”