Credit Union AI in Lending: Three Numbers Worth Tracking

Will Drewes

Founder, Fern Strategy · 2 min read

Credit Union AI in Lending: Three Numbers Worth Tracking

Most AI coverage in financial services leads with potential. These three credit unions logged actual results, in production, in regulated lending workflows. The numbers are specific enough to be useful.

FORUM Credit Union: 70% Faster Loan Processing

FORUM (Indiana) deployed AI-driven document processing to clear the bottleneck that kills most loan timelines: document verification. The result was a 70% reduction in processing time, with members receiving decisions in hours rather than days.

That gap matters operationally. Loan officers weren't slow; the queue was slow. Intelligent document processing doesn't replace underwriting judgment, it removes the manual extraction work that sits upstream of it. When that step compresses, the whole pipeline moves.

Centris Federal Credit Union: Auto Underwriting That Actually Scaled

Centris ran their auto loan book through AI underwriting and moved automated decision rates from 43% to 63% over 2024. Indirect lending volume grew more than 30% in the same period.

The part worth pausing on: AI-approved loans showed more reliable credit quality than traditional scoring models. That's a direct challenge to the assumption that automation means looser standards. At Centris, it meant more consistent application of the same standards, at higher volume, with less variance in outcomes.

For any credit union still treating AI underwriting as a future-state experiment, Centris ran this in production and measured it.

Community Service Credit Union: Personalization at the Right Moment

Community Service CU targeted members at financial milestones with AI-driven personalization. Within six months: 25% increase in lending acquisition, a deposit conversion rate 5.4 times higher than non-personalized benchmarks, and 35% lift in overall engagement.

The 5.4x deposit conversion figure is the one to focus on. Personalization in financial services usually gets measured in click rates. A conversion multiplier that high, on deposits, points to timing and relevance doing real work, not just better subject lines.

What This Means for Credit Union Leaders

These three cases sit in different parts of the lending workflow: document intake, underwriting decisions, and member acquisition. The outcomes are different in kind, not just in magnitude. That's useful because it means the entry point for AI in your operation doesn't have to be a single big bet.

The governance question is where most credit unions are still working things out. NCUA has published AI guidance, but audit trails, model explainability, and fair lending documentation are still operationally immature at most institutions. Centris can point to credit quality data that validates their model. Not every deployment has that.

Before you build or buy, the question worth asking is whether you can produce that kind of evidence on demand, for an examiner, for your board, or for a member who asks why they were declined. If the answer is uncertain, that's the gap to close first.