Credit Unions Turn to AI to Expand Lending Access

Credit unions are going all in on artificial intelligence to a degree that would have been unlikely a few years ago. Now two-thirds of credit unions indicate they will employ AI to make credit decisions.

That signals a new chapter for thin-file members that have long been unable to penetrate traditional scoring.

The tools continue to improve as members now require fast responses, and market competition continues to expand. A credit union that fails to meet expectations may prompt members to consider alternative institutions, including banks, fintechs, or technology companies that offer financial services. 

Wisconsin-based Marine Credit Union’s move to partnering with GDS Link is a case study in what a mid-size credit union does when it hits a wall. Kenneth Brossman, Chief Lending Officer, put it plainly: “We wanted to expand our auto decisioning percentage and get in that third-party data such as LexisNexis, and that’s where we needed help.”

That help came in ways the team didn’t expect. GDS Link wasn’t just a model builder. It ran backtesting, provided monthly monitoring, and gave Marine the paperwork regulators look for.

Steven LeJeune, Business Intelligence Manager at Marine Credit Union, explained why support for compliance made a difference: “A fair model that’s compliant is the biggest part of it, but also the back testing and continuing monitoring. GDS Link does all of that for us.”

By having these processes outsourced, Marine did not have to bear the cost of in-house documentation.

The system gave them what score auditors desired — strong evidence, constant monitoring, and evidence of provability.

To a credit union that is juggling sophisticated member demands, that guarantee is no less significant than scores themselves.

Today, Marine’s system makes a call on more than half its consumer loans, almost 57% in fact, with two-thirds resulting in approvals and one-third in denials. Then comes another layer: a second model that reexamines declines, designed to give subprime borrowers a fairer shake.

“We live in a day and age where everybody wants instant answers,” Brossman said. “We’re not only competing with the big banks, but with the Googles and Amazons of the world.”

That urgency is only part of it, however. Marine informs people of their situation in cases where a life story is more important than a score.

“We want people involved where that story really does matter … a human being can identify those things and tell that story for us,” said Brossman. It’s a reminder that automation has limits, and that the member relationship still matters. For credit unions, that hybrid model is quickly becoming the default.

The results have become apparent. The current delinquency rates are lower at Marine (though they remain higher than peer institutions) and charge-offs have decreased because of improved member selection processes. 

Brossman was blunt: “Not all credit scores are created equally. Being able to deploy information that’s not on a credit bureau report … might allow you to take a really good risk for a 580 borrower and not take a bad risk on a 680 borrower.”

The logic is simple. Add in rental, utility, or phone payment histories, and you can separate a member who’s building stability from one who isn’t.

LeJeune added, “If we’re able to auto approve or auto decline some of those on either side, but then spend the time with those others that we really need to get the story of, that helps us.”

Other credit unions are experiencing similar results. Indiana-based FORUM Credit Union reduced loan processing time by 70% through AI-assisted document review. Centris Federal Credit Union increased its auto loans by 43% to 63% within a year.

“Being able to deploy information that’s not on a credit bureau report… might allow you to take a really good risk for a 580 borrower and not take a bad risk on a 680 borrower.” — Ken Brossman, Chief Lending Officer at Marine CU

The trend is appearing throughout the movement: Automation combined with beefed-up risk perceptions is enabling thin-file borrowers to receive a better deal.

These gains come to credit unions confronted by increasing subprime pressure as unsecured debt balances and defaults are climbing. Brossman said rising defaults and unsecured debt are putting stress on credit unions that cater to higher risk markets.

Marine has emphasized restructuring alternatives, assisting members in combining revolving card debt into installment loans of shorter duration and lower interest rates.

LeJeune explained that coaching and debt counseling are included in the credit union’s approach, which help subprime families save money on interest and shorten collection times.

Marine is not only using AI for approving loans. They also introduced an AI phone system to handle routine questions and created a knowledge bank that reduced staff search time from 20 minutes to seconds.

Small victories such as these add up by putting people to work on tasks that call for judgment while eliminating bottlenecks.

Marine’s adopted tools for member use reveal a larger reality: AI is infiltrating corners of credit unions where inefficiency once dwelt. Questions of a routine nature are answered instantly. Staff spend less time searching and more time making informed decisions. These advances contribute to how members perceive service. 

Marine’s next move is to create a data lakehouse through Snowflake and Databricks to bring fragmented information into a unified zone. Lenders are also monitoring stablecoins and how they can impact deposits as well as collateral. Credit unions are using AI to keep up with pressure to enhance services.