ORLANDO, Fla.–An overflow crowd turned out here to hear more about how artificial intelligence can grow lending portfolios.

Speaking to the League of Credit Unions’ ENGAGE Conference, Samantha Hubbard, VP-business development with Scienaptic, which provides AI-based lending decisioning, said that when discussing AI in the context of credit union lending she is talking about predictive analysis—taking a lot of data and analyzing it quickly—and not generative AI such as ChatGPT.

Chasing efficiencies and just looking to compete in a difficult lending market, Hubbard cited data showing the rise in delinquencies, a much slower loan market, and a 4.7% reduction in net income that is putting pressure on margines as among the reasons CUs have turned to AI.

Hurry Up & Decide

All of that is coming at the same time members want decisions made more quickly than ever, she said, pointing to a survey that found 74% of members expect loan decision in less than 24 hours. Moreover, a third of borrowers have thin files or no credit history, and one in three loan rejections are due to outdated decision rules, she said. 

Scienaptic has 70 plus credit unions using its AI-based solutions currently, and is currently approving $3 billion in loan decisions monthly, she said.

“AI isn’t replacing the heart of credit union lending, it’s amplifying it. It’s helping them to do it better and more efficiently,” Hubbard said. “It’s about how can we help the membership and make it a better experience.”

Sitting with a loan office for two hours may show an investment in the member by the CU, Hubbard observed, but what that member really want is to get a decision made quickly. 

Signals Monitored in Milliseconds

AI can evaluate 3,000-plus signals in milliseconds, according to Hubbard, and that includes significant data points from LexisNexix, as well as all sorts of other behavioral data, which is especially critical with thin and no-file borrowers.

Scienaptic is reporting 90% or more of those without a traditional credit history can be scored on its platforms. It said that figure is 45% among protected classes, such as African-Americans, women, people older than 62, Hispanics and Asians.

AI-assisted loan decisions, said Hubbard, “is not a reason to reject people. The alternative data is to get to more yesses. And this is without increasing risk. That is key. You can get 25% to 40% more approvals without increasing risk.”

Life Lessons

Hubbard said credit unions on the Scienaptic platform saw 3.8% loan growth from December 20234 to December 2024.

She further noted the data is also be trained on real-word events that have occurred, such as the 2008-09 recession, hurricanes and the California wildfires, all of which, Hubbard suggested, leads to “recession resistant lending.”

Other Topics

Other points made by Hubbard:

Job Losses

Responding to a common concern, Hubbard said no loan officers lose their jobs as a result of AI.

ROI

Hubbard said ROI varies by asset size, according to Hubbard: Scienaptic’s experience shows:

• 8-15x for $500 million-and below in assets
• 11-23x for $500 million-$1 billion in assets
• 15-35x for CUs of assets of more than $1 billion

Loan Denials

Hubbard said some CUs don’t want to auto-deny any borrower. They want the loan officer to review and then contact the borrower. Other CUs will auto-decline, and in those cases the reasons for the adverse-action is provided.

Compliance

Hubbard said all of the company’s client credit unions have “aced” their NCUA exams.