Simple Subscribe
Subscribe Now!
Most community banks and credit unions are aware by now that their workforces (to say nothing of their leadership teams) comprise both advocates and resisters when it comes to artificial intelligence.
Many of these institutions likely see the tension as a source of strength, the kind of viewpoint diversity that leads to better decision-making and outcomes. Others may see it more straightforwardly as an obstacle to progress.
Either way, with commercial LLMs now three years old, the advocate/resister divide — especially among smaller financial institutions — is a fact of life and a challenge to be reckoned with.
And it’s gaining urgency as the competitive baseline shifts. Large banks and fintechs are already using AI to cut costs, improve risk decisions and reduce fraud, and improve customer acquisition and user experience. Lagging institutions, including small and mid-sized ones, will face higher cost structures on a relative basis and have a harder time winning and retaining customer relationships.
The job to be done is to get the two camps communicating — constructively working through their areas of doubt and disagreement, selling one another on the benefits of moving forward, and firming up the use cases that will deliver the highest value with the least risk.
To learn more, we spoke with Vivek Sanghvi, Sales Engineer at Narmi, which specializes in helping small financial institutions level up their technology capabilities. He offered insights on the mindset shifts, proof points, and use cases that can bridge the gap and move institutions forward.
Want to read more like this? Check out Narmi’s content portal on The Financial Brand: Be Where Banking is Going
Mindsets and Mindset Shifts
Don’t assume AI fear is rooted in deeply held principles. Most often it is a reflexive response to a worst-case scenario extrapolated from press reports or simply to the unknown. For example, compliance teams might envision autonomous systems making consequential decisions without oversight — something that very few (if any) institutions would propose. A good place for the advocate to start is by closing the gap between perception and reality through education.
“Compliance officers know a lot about bank regulations and risk management, but the specific mechanics of AI deployment are often entirely new territory,” Sanghvi said.
Stop painting with a broad brush. Even with all of the media attention, many businesspeople haven’t yet grokked the difference between generative and agentic AI — something most AI advocates might not realize. The distinction matters more in bank and credit union environments, where anything in autopilot mode is anathema. (The difference, for those readers who might be “asking for a friend,” is that generative AI responds only to a user’s prompts, whereas agentic AI can trigger and complete tasks on its own initiative.)
Sitting down with decision-makers, perhaps even board members, to define basic terms, may allow everyone to restart conversations on a less adversarial basis.
Clear the air about layoffs. Some bank and credit union executives might too eagerly anticipate immediate AI-driven cost reductions. Others may consider the potential for mass layoffs a red flag, because they’re concerned about talent loss and, again, the prospect of critical functions on autopilot.
The better message is this: AI will deliver efficiencies but likely first as collaborator, not a replacement, of humans. “I like to describe AI as a ‘smart intern,'” Sanghvi said. “AI will complete eighty or eighty-five percent of the task, but you’re still going to double-check that intern’s work, making sure all of the details are in order.”
Understand the cultural fault lines. Employee and C-suite anxiety often run counter to each other. “It’s very generational,” Sanghvi said. “Effective advocates will address each group on its own terms.” Junior staff, frequently enthusiastic about automation and the technology itself, worry about their jobs.
Senior leadership, seeking business upsides but less certain about the technology itself, may be more vocal about risk. (To the degree they are worried about their jobs, they’re likelier to keep it to themselves.) A demo or presentation that doesn’t account for both fear profiles will land wrong with at least part of the room.
Tactics that Drive Adoption
Anchor in the institution’s pain points. Sanghvi recommends that AI advocates undertake a discovery process focused entirely on identifying pre-existing problems that AI can solve. These might include high call-center volume; or consumer-facing workflows that still require branch visits; or staffers who wear multiple hats, one of which invariably requires performing repetitive, error-prone tasks. When you host a vendor, have them focus exclusively on functionality that directly addresses such concerns.
Resist the urge to wow the team with a system’s breadth and depth. “If you’re just throwing the whole kitchen sink demo at them, showing them every nook and cranny… it could get very overwhelming.”
Don’t advocate for AI, advocate for your customers and members. At a time when competition for new accounts is fierce, and the battle for primacy even more so, AI can be a force multiplier. Internal advocates would do well to focus on AI’s potential to improve the customer or member experience. Start by identifying the customer and accountholder personas with the highest likelihood to benefit from and embrace AI-assisted products and services.
A useful decision matrix might score accountholders and prospects along two axes — the value they would gain from an AI capability and their receptivity to AI — and focus first on serving those in the upper-right quadrant. Institutions wary of moving too fast can observe AI’s impact within a contained population before committing to a wider rollout.
Establish a “crawl, walk, run” roadmap. This will reassure both resisters and those eager for fast gains, showing the former that nothing advances without leadership approval and the latter that everything is possible in due course. Here’s a top-line view of what that might look like:
- Crawl: pick 1–2 low-risk use cases, write AI usage rules, and train staff on approved tools and data handling.
- Walk: run small pilots in areas like fraud, compliance, customer service, or internal drafting, with human review and clear success metrics.
- Run: scale the pilots that worked, integrate them into core workflows, and tie them to governance, model risk, and board oversight.
Applications and Use Cases
For organizations that are ready to lean into AI but unsure where to start, here are some of the applications and use cases that financial institutions have been benefiting from.
AI-assisted secure messaging: Call center volume is among the most cited pain points at community institutions, and one of the most fixable. When a member sends a secure message through the platform, a staff member prompts the AI to draft a response. The draft appears in seconds; the staffer edits, approves, and sends. For employees already handling multiple roles across a single workday, the time recovered is meaningful without any reduction in service quality.
Promotional card generation: Generating basic creative assets for marketing can be a chokepoint for small institutions. And the requirement for agility in marketing is on the rise. Creating a promotional asset for a new CD rate can consume hours, even with tools like Canva. AI-driven tools, instructed by plain-English text prompts, can produce formatted promotional cards in a couple of minutes. Using a live preview function, a staffer can then edit and approve directly. The capability adds scale and reduces time-to-market without adding headcount.
KYC/KYB decisioning: According to Sanghvi, AI skeptics often sit up and take notice when presented with fraud-reduction applications. Walking a skeptic through the decisioning logic behind a KYC/KYB engine reframes the conversation, shifting AI from a defensive posture (driver of risk) to an offensive one (neutralizer of risk). Narmi’s platform, for example, can evaluate hundreds of inputs simultaneously during onboarding, Sanghvi said — including straightforward data points like IP address and geolocation and esoteric ones like device jailbreak status. The work is clearly faster, more comprehensive, and more auditable than current standards. “Their eyes will open up like, whoa… this is a whole other world,” Sanghvi said.
MCP server: Financial institutions, enabled by technology partners, are increasingly offering their customers and members the ability to query their financial data through tools like ChatGPT and Claude. MCP server enables read-only access to live banking data, queryable in plain language. This capability will likely appeal to current and potential accountholders in the high benefit/high receptivity quadrant of an institution’s AI persona matrix, including younger consumers and more tech-forward businesses.
Most institutions today will talk about the necessity of establishing ground rules before advancing down the path of AI deployment. But use cases like these are what will ultimately give your advocates and resisters the firmest common ground to stand on — and put your institution in a position to move into the future on a competitive footing.