Credit Union CEOs: Choosing the Right AI — Not Just More AI

As the former chief executive of Vantage West Credit Union, I spent more than two decades watching technology vendors beat a path to our door. During my tenure we transformed a small military credit union into a member-owned financial institution with 150,000 members and $1.9 billion in assets, making Vantage West the third-largest credit union in Arizona. Those numbers matter because they represent the trust our members placed in us. They also gave me a front-row seat to the promises — and pitfalls — of "AI-powered transformation."

Today, credit union CEOs are bombarded with AI solutions that claim to improve member experience, cut costs and accelerate loan growth. Many of these tools are well-intentioned, and there is no question that AI holds promise. The National Credit Union Administration (NCUA) notes that credit unions are exploring AI to enhance member services and streamline operations, but the agency warns that the technology also introduces unique risks. Before approving another pilot, we must be honest about why so many AI initiatives fail — and what it takes to succeed in a regulated, trust-based environment.

The Real Reasons AI Pilots Fail

Most failed AI projects aren't the result of bad algorithms; they're the product of bad integration. In my experience, three factors doom more pilots than any technical bug.

• ·Lack of integration into the core. Tools that sit on top of existing workflows without integrating into core banking systems simply add friction. They generate duplicate work or force staff to perform manual workarounds. If your team has to "work around" the AI, you're incurring additional costs rather than reducing them.
• ·Governance and compliance gaps. The NCUA points out that AI adoption raises questions about algorithmic decision-making, fair lending compliance, member data privacy and operational resilience. Credit unions operate under strict regulatory oversight; we can't afford to deploy opaque models that introduce bias or expose sensitive data. Any AI solution must come with clear risk management, audit trails and a commitment to human oversight.
• ·Absence of measurable outcomes. Too often pilots are green-lit on the promise of efficiency or "member experience" without tying those benefits to the balance sheet. In a credit union, ROA, member growth and net loan growth are the metrics that matter. If a vendor can't articulate how their tool improves one of those metrics or reduces risk, it's not ready for your institution.

Five Questions Every CEO Should Ask Before Signing a Contract

Before you pilot another AI tool, ask these five uncomfortable questions. They may save your organization millions — and preserve the trust you've worked so hard to earn.

• ·How does this impact the balance sheet? In credit unions, we measure success in loan growth, servicing costs and risk exposure. A tool that boosts member satisfaction but doesn't move these levers is a nice-to-have, not a must-have.
• ·What risks am I taking on? Beyond vendor-provided assurances, insist on a clear description of how the model handles member data, protects privacy, mitigates bias in credit decisions and supports regulatory compliance.
• ·Is this replacing work or adding complexity? AI should free your team to focus on high-value conversations, not add another interface for them to manage. If the "automation" requires staff to babysit or double-check the output, you're layering complexity onto already stretched teams.
• ·What happens in year two? Many pilots are priced attractively in year one, only to see pricing surge after you're locked in. Understand the long-term cost of model maintenance, vendor lock-in, and ongoing training — especially if your institution is responsible for model retraining or compliance updates.
• ·Can they demonstrate success in a credit union like yours? Success stories from fintech startups or mega-banks don't translate to member-owned institutions. Require evidence from organizations of similar size and regulatory complexity. Our mission and risk tolerance differ from shareholder-driven institutions.

The path to AI success is not to adopt the most AI; it is to adopt the right AI, in the right context, with the right controls.

Advice to AI Vendors

For AI vendors seeking to serve credit unions, your greatest competitor is skepticism. We've heard years of "digital transformation" pitches. To earn our attention:

• ·Stop emphasizing AI for its own sake. We assume your product uses AI. Tell us instead which specific operational problem you solve — whether it's reducing fraud losses, improving loan underwriting accuracy, or decreasing member service wait times.
• ·Translate your value into risk reduction and balance-sheet impact. Show how your tool will lower loan delinquency rates, increase cross-sell opportunities, or improve capital ratios. Tie your success metrics to ours.
• ·Provide evidence. Demos and polished decks aren't enough. We need case studies from regulated, member-first institutions. If you don't have them, partner with a credit union to build one before coming to market.
• ·Address compliance and governance up front. Vendors must arrive with audit logs, explainability and human-override mechanisms baked into product design — not bolted on later.

Trust, Clarity and Choosing the Right AI

Credit unions are built on trust. Members choose us because we put people before profits, and we can never compromise that trust for the sake of shiny technology. Clarity on business impact and risk doesn't slow down innovation — it accelerates decisions by reducing uncertainty. As CEOs, our job is to protect the cooperative and unlock value for members. Anything less than thoughtful, integrated AI adoption is just costly experimentation in an industry that does not tolerate errors.