PALM DESERT, Calif.–The power of artificial intelligence (AI) and machine learning is not just about the numbers when it comes to lending, it’s also about fulfilling the core credit union mission and expanding markets, according to one person.
Speaking to the California and Nevada CU Leagues’ REACH Conference here, Mike de Vere, president of Zest AI, told the meeting, “The current credit system is failing America. You can see it when you look at the disparity between FICO scores for female borrowers and male borrowers. Female borrowers’ scores are 40 points lower on average, but their debt is also 18% lower on average or lower. The model used today was created in the 1950s.”
The answer to removing the kinds of bias that lead to those disparities can be found in AI, de Vere said, stating automated underwriting is more accurate and inclusive.
The demand is there according to de Vere, who cited a survey showing 98% of CU executives say they are looking to leverage AI or machine learning, while 80% of members want their credit union to use modern technology to assess their credit worthiness.
“But it’s not easy,” he said. “You’ve got to build the model, solve for explaining how the model was built, generate adverse action codes that are validated, conduct fair lending testing and a method to de-bias, create compliance documentation, and you must monitor AI while in production.”
Broader Pool
But in doing all that, he said, the immediate benefit is access to a much broader pool of borrowers.
“One-in-five members don’t fit the credit system; that’s 46 million Americans,” de Vere said. “The legacy approach is unable to score the underbanked. We can do better by consuming more data and applying more math. The best part of AI and machine learning is it can apply the missing math. You can build out scores for 98% of your members.
“The second area is around expanding access to credit. That’s a tall one,” he continued. “It’s a top priority among credit union members, with 78% saying they want expanded access to credit. In times of crisis, credit unions often shrink the credit box and deal only with A paper members. Is that living your mission? What’s really cool is that credit union members love you and they look to you for help. How do we do that? With an underwriting process powered by AI.”
de Vere said in the case of the AI employed by Zest AI, it starts with the data a credit union already has.
“It’s not about scraping social media accounts. It’s about a few hundred or so variables. FICO boils people down to a three-digit numbers. People are more than a number. There are borrowers you can swap in, but also borrowers you can swap out,” he said. “You can see a huge lift over benchmark.”
That huge life includes across the credit tiers a boost in loan approvals by 39% overall. There is even an average 8% increase in loans to members with grade A credit scores, he said.
Underserving the Underserved?
“Often, when credit unions talk about serving their community, they are really serving their top tier,” said de Vere. “The other benefit is around consistency and inclusion. Whether you are big or small, when we look across your underwriters and they get a file put in front to them, will it be a consistent decision? Of course not. Bias exists in the credit system. We need to tune down proxies for race or gender.
“And in this new world members want quicker decisions,” de Vere continued. “Eight percent of credit union executives in survey said it takes more than 30 minutes for a decision. Do you think that is members expectations? Seventy-two percent of members want a decision in seconds. So, if you want to create a thriving, durable credit union, expectations are going up.”