Written by: Shea Gabrielleschi
Everyone seems to be endlessly curious how everyone else is implementing AI. If you are wondering where your community bank peers actually are, most are not racing ahead with radical transformation. Most are cautiously experimenting, building guardrails, and trying to identify practical use cases that tie directly to business outcomes.
Over the past couple of years, I have had more conversations about AI with bank CEOs, COOs and CIOs than I care to admit. Some of those conversations have been private. Others have taken place in peer groups where bank leaders are candid about what they are trying, what is working, and what is frustrating.
Below I share what I've learned from my conversations, particularly in recent months.
"Give Me a Real Use Case."
Boards are asking about AI. Executive teams are discussing it. Committees are forming. Vendor demos are happening. Policies are being drafted. Despite all of that activity, there is a frustrated tone I sense in my conversations with CEOs.
Bankers are tired of being told they need to adopt AI. After the 254th webinar and breakout session, they know that. It's time to execute, but most are stuck in a holding pattern looking for use cases.
One COO told me last summer, “Give me something practical that makes one team more efficient this quarter.” I hope he's found three since we last spoke!
The proactive banks are surveying department heads for pain points that automation could address. They are building lists of potential use cases. They are narrowing those lists down to a handful of controlled pilots. They are asking whether AI can support underwriting workflows, quality control reviews, call center queue management, internal policy navigation, or employee training initiatives.
Even with structure in place, the process is not simple, though.
A CEO of an $800M bank recently said to me, “We know this matters. We just don’t want to get it wrong.”
A CIO of a community bank in Pennsylvania made an observation that I think is important: He said that AI can make an individual task faster, but that does not automatically improve the entire process. If one person drafts something more quickly but the downstream workflow remains unchanged, the net impact may be modest. I think this is why you don't see AI bringing radical efficiencies and changes overnight yet.
Where Is AI Actually Showing Up?
When I step back from individual conversations, I see AI showing up inside community banks in four primary ways.
1. First, through Microsoft Copilot inside the Microsoft 365 environment. When a bank CEO says they are "starting to dabble with AI," usually they are talking about Copilot. Copilot has become the default entry point because it lives inside a tenant that 90% banks already trust and integrates with Outlook, Teams, and SharePoint. It feels contained and manageable.
Most banks are starting with Copilot Chat through existing licensing. Fewer have expanded into broader paid licenses. One COO told me about 4 months ago, “We do not think the ROI is there yet to pay for broader licensing.” With recent improvements to the model, I'm curious now if that COO has found the ROI.
Usage is rising, and in some banks a majority of employees have interacted with Copilot in some way. Much of that usage is still light, summarizing emails or generating draft content rather than deeply embedded workflow automation. Leaders are watching closely to see when incremental productivity gains translate into measurable economic impact.
2. Second, AI is being consumed through enhanced features inside familiar third-party vendors. Fraud and risk platforms are layering in machine learning enhancements, like Ncontracts' Contract Ntelligence Solution within Nvendor. CRM systems are embedding predictive capabilities, like Creatio's AI-native CRM. Data analytics tools like Quantalytix, KlariVis , Revio Insight, and others are all using AI to become more user friendly and API-friendly. Meanwhile, the big core providers are just trying to keep up.
3. Third, banks are implementing purpose-built AI fintech solutions aimed at specific operational gaps. A few have explored internal knowledge bots tied to SharePoint libraries or policy repositories, sometimes in partnership with vendors like Posh and Hapax or through in-house experimentation. Others are evaluating workflow automation tools such as HuLoop Automation and DeepSee.ai to address process gaps.
How are your peers finding these fintech solutions? Through trusted places all over. Some are plugged into fintech ecosystems like BankTech Ventures and Alloy Labs. Some are going to ICBA, American Bankers Association and their state associations for recommendations. Thought leaders like Bank Director and American Banker put forth a good effort to parse through all the AI clutter for banks.
4. Fourth, there is the rogue usage that sits outside of the strategy. Even in banks that have not broadly deployed AI tools, employees are using ChatGPT on their own. I have heard examples of lenders pasting in rough credit memos to improve structure and clarity, marketing teams drafting campaign copy, and operations staff summarizing long policy documents to save time. In most cases, the goal is simple productivity.
One executive told me, “We are not going to totally shut this down, because it is going to happen. We would rather keep our eyes on how it is happening and bring those capabilities into our environment.” I love that statement.
Hidden usage, known as shadow AI, is already inside most institutions. Since you can't just shut it down, make sure you're giving your employees safe tools that are also effective.
If the "allowed safe" tools don't hold a candle to the "forbidden better" tools, you're battling human incentives and accepting the reality that many of your top employees are using prohibited tools to work smarter and faster.
This is why it's critical to find the "best safe" tools, and implement this quickly with proper governance.
Across all four categories, most of what I see is incremental rather than transformational...for now. The emphasis remains on internal efficiency, improved clarity, and manageable experimentation.
You could make an argument a fifth category is internal development of AI agents and workflow bots, but I'm just not seeing enough banks doing this on their own to earn a full spot yet.
Who Owns AI Strategy and Execution Right Now?
The billion dollar question... When I ask banks who owns AI, the answer is usually a committee. In many banks, AI has become shared territory across IT, risk, compliance, legal, fraud, and operations, because each group sees a different part of the exposure.
When I talk to the banks making progress implementing AI, that committee group has a clear mandate and operation. They run an intake process for AI ideas, evaluate risk early, pick a small number of use cases worth testing, and track what comes out the other side.
Carey Ransom of BankTech Ventures recently shared this comment on Linkedin:
"AI strategy has to be part of business strategy. If it’s stuck in IT it will never be as impactful as it can or should be."
That comment is 100% true to what I see playing out in real banks. AI gets traction when it is treated like a business capability, with clear ownership. The committee structure works best when it has a simple mandate, a lightweight intake process, and the authority to say yes or no quickly. Some CEOs have complained to me that their AI committees have slipped into endless discussion with no outcomes.
The biggest takeaway for me is that AI strategy is becoming inseparable from business strategy, and when business lines bring the problems and IT brings the structure, things get done. Easier said than done, but that is what's happening.
Related: Cutting Fallen Giants and IPOs Transforms Small Cap Performance