Over the past six months, I’ve had the pleasure to speak with approximately 50 financial institution (“FI”) CEOs and senior executives about AI. These weren’t panel discussions or surveys — they were 1:1 conversations with Narmi clients and prospects in FI offices, Narmi’s corporate headquarters and conferences.
The tone of these meetings was candid, sometimes skeptical, a bit overwhelmed and largely excited. What follows are a few themes I keep coming back to and have relayed to my teams internally:
The institutions making the most progress on AI are not the ones trying to implement it across every function at once. They’re the ones who picked one or two high-impact use cases, moved thoughtfully, and built internal confidence before expanding.
The “AI Everywhere” narrative — pushed hard by some — creates a kind of paralysis. Leaders feel behind before they’ve started. It doesn’t help that there’s a visible gap emerging between a small group of innovative, fast-moving institutions already running multiple AI features in production, and the majority who are still finding their footing. That gap is real, but it’s not insurmountable — and trying to close it by doing everything at once is the wrong response.
The antidote is focus: identify where AI can have the clearest, most demonstrable impact, go deep there first, and let success in one area build momentum for the next.
This was the most consistent theme across every conversation. As of right now, community FI leaders are not looking for AI to remove humans from the loop — especially not in decisions that affect end-users. What they want is for AI to make their people faster and better-informed.
The framing that resonated most: “Help my staff make the best decisions quickly.” Not “automate the decision.” Not “replace the reviewer.” Surface the right signals so the human can act with confidence and speed.
We see this in real-life. The AI features getting the most traction at Narmi are ones that augment existing workflows — drafting a message, surfacing a KYC data point, generating a marketing audience — rather than ones that attempt to fully automate a process end to end. Perhaps we will get there eventually, but 100% “replace the human” is not the focus right now.
FIs are not going to hand decisions to a black box. Full stop. The institutions most open to AI are the ones where they can see why the system made a suggestion — not just what it suggested.
This came up repeatedly: confidence scores matter less than reasoning breakdowns. Leaders want an audit trail — what did the AI look at, in what order, and why did it land where it did? One participant described explainability as a valuable exercise “even before any automation kicks in.” The process of understanding how the AI thinks is itself trust-building.
It’s worth noting that signing Narmi’s AI Terms and Conditions doesn’t automatically confer confidence. We’ve seen FIs who have technically cleared the governance hurdle still hesitate to activate features because they don’t yet have a clear mental model of how the AI works or who’s accountable when it doesn’t. Paperwork and trust are not the same thing.
For vendors trying to implement AI at financial institutions, this is a design mandate, not a nice-to-have. Logging, transparency, and interpretability need to be first-class features, not phase two items.
Nobody I’ve spoken with is opposed to AI eventually acting autonomously in well-defined scenarios. Skeptical? Definitely. Opposed? No. But the path there matters enormously. The framing that lands consistently: you earn autonomy by demonstrating accuracy over time, in specific buckets, and then letting the institution flip the switch themselves.
FIs want to be in control of when and where automation kicks in. The vendors who get this build incrementally — starting with assist mode, proving the model, and only expanding scope as confidence is established. The vendors who don’t get this tend to over-promise and underdeliver.
Here’s a pattern that comes up more than almost anything else: a forward-thinking operations or product leader at an FI is bought in. They understand the value. They want to move. And then the initiative stalls — not because of budget, not because of technology, but because it’s sitting in a Legal or Compliance queue with no clear timeline.
This isn’t a criticism of Legal and Compliance teams — their caution is legitimate and their job is hard. But there’s a real gap between signing AI terms and conditions and actually feeling confident about what those terms mean in practice. Many FIs have technically cleared the governance hurdle on paper while still carrying what I’d call “AI frontier anxiety” — a lingering unease about liability, data access, and what happens when something goes wrong.
The FIs making the most progress have found a way to bring Legal and Compliance into the conversation early — not as a gatekeeper at the end of the process, but as an informed partner from the start. That means vendors need to be able to speak fluently to compliance concerns, not just product benefits. And it means FI leaders need to invest in building internal literacy around AI policy, not just AI capability.
Up until about a year ago, AI sat comfortably on the CTO’s desk. It was a technology question: what tools do we evaluate, what vendors do we pilot, what integrations do we build? What’s our AI policy? CEOs could stay at arm’s length.
That’s no longer true. Now that AI directly correlates to human efficiency — how many people you need, what they spend their time on, how productive each person can be and how much we pay them — it has moved squarely into CEO territory. Headcount planning, org design, and employee productivity are CEO decisions. And increasingly, AI is the variable that shapes all three.
The CEOs I’ve spoken with who are furthest ahead on AI are the ones who claimed ownership of it early — not by becoming technical experts, but by treating it as a strategic and operational priority that sits on their desk, not just their CTO’s. The ones who are behind tend to still be waiting for a technology recommendation to bubble up.
The community and regional FIs that will lead on AI over the next few years aren’t the ones with the biggest budgets. They’re the ones building the right governance foundations, being thoughtful about where to start, and choosing partners who understand that trust is the core product.
If you’re a FI leader thinking through your AI strategy and want to compare notes, please do reach out — we’re always interested in trading notes and contributing to the financial institution community.
.jpg)
