The Informed Team is just back from a productive Fintech Nexus USA (formerly Lendit Fintech) where our CEO Justin Wickett was joined onstage by Mike Qin, Vice President, Business Analysis at Capital One for a conversation moderated by John Siracusa, Host and Producer of the “Bank On It” podcast.
Here are some excerpts from their conversation:
John: The auto industry is full of opaqueness preventing consumers from seeing what is happening behind the scenes. My question is how can AI and technology fix that?
Justin: Opaqueness is definitely a pain point. When we talk to borrowers applying for credit, they say they spent hours at the dealership trying to figure out what they qualify for. We found in user interviews that it’s not just the consumer who is left in the dark. The staff at the dealership isn’t sure how the lender will account for overtime pay, bonuses or tips that a lot of these borrowers have. When we analyze pay stubs, we see over 50% of these borrowers have variable pay.
Mike: Many of you know Capital One as a big player in the credit card market. We actually have a sizable auto finance portfolio as well – We’re the largest non-capital lender in the US. It is evident how many friction points a customer has to go through to secure an auto loan. I lead a team which verifies the identity of applicants. We look at the borrower, as well as their ability to repay, which comes from formal verification of employment, income and interest. The collection of all of these documents from a network of ~17,000 dealerships creates many degrees of risk, a lot of friction points for the customer.
Justin: And it’s not just opaqueness. The current process is error prone and subject to bias. We hear about error rates of 10 to 15%. As it relates to bias we’ve had lenders tell us that they gave loan jackets to 10 different loan officers on staff and got 10 different income calculations. And that’s not uncommon. There’s also a lot of subjectivity that results in matters requiring attention from the lenders’ regulators. AI brings a systemic approach to calculating applicant income and does so in real time without having to expose data outside of your organization.
Mike: I agree absolutely. I just did a financial exercise with my team where I got a bunch of real pay stubs and asked them to calculate the income. I think we did it with eight or nine people. Not a single person got the same answer as someone else. So just that depth of subjectivity manifested through human execution, you see inconsistencies.
Mike: Justin talked about a speed element which is really important and consumers are expecting that more and more. They don’t want to be sitting in the dealership or loan office for hours waiting for you to run a bunch of validations and checks.They want the speed you can’t get without AI. The customer may be sitting in the office late Saturday night finalizing the structure of the payments. And now you say hey, do you have that piece? If we’re unable to scan that right away and give the dealer an answer, the dealer may be left holding up the customer from driving away in their brand new vehicle. They really don’t want to have to say, “come back on Monday.” So having the technology upfront creates a lot of speed and erases friction.
John: Mike, I’m curious. The F&I manager at a dealership selects which lender they want to work with, right? Do you feel you get more opportunities by having a better process?
Mike: Totally. The auto market is very much like an auction house. The dealership selects which lender they want to send a contract to. You essentially compete on price, experience and your relationship with the lender. So, having a product that has the AI embedded, where we are able to deliver the promises of speed and confidence, definitely helps the dealer choose us as the lender.
John: And Justin, have you noticed that as well across your spectrum of clients?
Justin: Absolutely. We see lenders embracing AI to digitally transform their origination process. It helps drive improvements in efficiency and capture rate so they’re able to board more loans. For example, auto lending is underwritten based on stated income for the most part. When lenders verify income, on average about 30% of these full spectrum consumers applying for credit – we’re talking about truck drivers, construction workers – will understate their income by, on average, $7,000. So these people are in some cases qualified for access to credit but at a much higher rate. The ability to leverage AI to verify applicant income improves capture rate and frees up staff to focus on other important functions.
John: Looking at current events, the pandemic really sped up digital adoption – everybody needed to do it, people couldn’t go into a physical location. So timelines for implementing new software and processes went from five years down to a few months. Do you find that’s happened in lending?
Mike: The pandemic has acted as a catalyst for a lot of digital adoption, especially in the car industry. The reception to digital products on the retail side and the financing side has accelerated. Prior to the pandemic we had seen the industry use of e-contracts at about 15% and that ability had been in place a long time. It was just very slow adoption. In the last two years it went up to 50%. Now a lot of dealers say, “ Well, we really need new solutions and frankly customers are saying, I don’t want to be sitting in a showroom for hours where I may be exposed to COVID.”
Justin: What we’ve seen in the industry is threefold, One, auto lenders want to do straight through processing. They want to automate verification of income, identity, residence and as much as possible. Two, a lot more applicants are applying online to prequalify for auto and other loans which has increased the amount of fraud. Some lenders have fraud rates of more than 10%. Imagine 10% of the people applying on your website or mobile app, submitting falsified income documents! So that’s a huge liability and a great opportunity to take advantage of AI to automate fraud detection. The third observation that we’ve seen during this pandemic as digital transformation has really taken a hold is the need for speed. One of our FI customers told us they previously outsourced manual data entry for their income documents. People were keying in information and it took 15 to 30 minutes to get a response. In a digital world consumers want answers in real time. Otherwise, they’re going elsewhere. We can deliver calculated incomes in accordance with Equal Credit Opportunity Act fair lending laws in a matter of seconds.
John: You talk a lot about AI. Are your AI efforts mostly focused on decisioning or is it also documentation and data automation?
Justin: Extracting information from documents is a painful process. It’s a perfect application for automation. We work with a contributory database where we’ve got over 15 million documents that we’ve processed on behalf of lenders. We started off solving a core problem which is a customer (who may have overtime pay on their pay stub) applying for a loan. We’re enabling consumers to have a more streamlined, frankly more humane process of getting a loan and ensuring they’re qualified for the best rate possible, whereas in the past, certain income wasn’t counted. Our goal is to bring a more streamlined process to our lenders and to drive efficiencies.
John: Justin, Mike, thank you so much, this has been quite informative.
Adine Deford is the VP of Marketing at Informed.IQ. She has more than 25 years of technology marketing experience serving industry leaders, world class marketing agencies and technology start-ups.