Numerica Credit Union will use alternative credit scoring solution Scienaptic AI to automate credit risk assessment for its treasury services.

The $3 billion Spokane Valley, Wash.-based credit union offers treasury management services, such as ACH, remote deposit capture, wires, and the anti-fraud cash management service Positive Pay, said Ryan Bernard, vice president of Commercial Credit and Operations at Numerica.

“Numerica’s commercial portfolio is pretty robust,” Bernard told Bank Automation News. “Utilizing Scienaptic AI will allow us to better assess some of the risk and some of those treasury management functions.”

Earlier this week, the $4.4 billion Richland, Wash.-based Gesa Credit Union also revealed it has implemented the artificial intelligence (AI)-driven solution to automate more than 50% of its auto loan applications, with approval rates increasing up to 20%, according to a Scienaptic AI spokesperson. So far, the New York-based data and analytics provider has about 15 clients in the Americas, Eric Steinhoff, Scienaptic executive vice president for client impact, told BAN.

Numerica has already deployed the solution to automate underwriting in its lending products and services, including consumer and commercial underwriting. It will also use the tool to automate credit card increases, preapprovals and product recommendations for its members.

The credit union began its rollout of the solution in July after working with 10 to 12 solution providers during the past 15 to 18 months, Bernard said. Other providers who offer alternative risk assessment include Zest AI, as well as the credit bureaus, said Steinhoff.

CEO Pankaj Kulshreshtha said he founded Scienaptic AI after he was declined credit due to high utilization — even though he was never in default. He wanted to create a way for banks to consider additional data in the underwriting process, he said.

“What our platform does is it enables all the lenders to bring in a lot of data into the decision-making,” Kulshreshtha said. “I read somewhere, actually, that there are only about 30% of Americans that are actually defaulted at any point time, that they’ve gone late on a particular payment, and so forth. But most companies, if you see their approval rate, very low rates are like somebody has 40%, 45%, 50% approval rates — for a loan product that’s supposed to be very, very good.”

Most Scienaptic AI customers see credit approval ratings of 20% to 30%, which means creditors assume they are unlikely to pay, Kulshreshtha said. That also means creditors are rejecting a larger population than necessary, he pointed out.

Scienaptic AI changes the credit risk calculation by using machine language and alternative data sources that identify relevant factors, such as what property the person owns and where; has the person frequently changed addresses; and opt-in data, such as whether they pay utilities and other bills on time. The company offers pre-trained algorithms for smaller customers and custom-built solutions based on historical data for its larger clients with long lending histories, such as Numerica.

The 160,049-member Numerica was founded in 1937 to support railway workers and employees of other transportation systems. It will leverage AI in its lending as part of an effort to reach the underbanked while increasing the credit union’s number of loans, Bernard said.

“The underbanked and underserved piece really aligns with Numerica’s core purpose of enhancing lives, fulfilling dreams and building communities,” Bernard said. “Using AI is going to allow us to serve more of our members in a better fashion within. Using Scienaptic AI also, we get to do it with very little to no risk to the credit union.”

Scienaptic AI has raised $9 million in funding over four rounds since its 2014 founding.

The Bank Automation News webinar on automation technology for exceptional bank cybersecurity and ID verification takes place on Thursday, Sept. 9, at 11:30 a.m. ET. Register here. Attendees will be able to ask questions via chat.