Everyone’s talking about AI—but for community banks, the real work starts behind the scenes. We sat down with KlariVis, ICBA ThinkTECH alumnus and Preferred Service Provider, to talk about data, decision-making, and what banks need to do now to be ready for what’s next.
Q1: Tell us a little about your background as a community banker and what prompted you to found KlariVis.
I started my career as a CPA at KPMG and in 2005 became the CFO of a publicly traded community bank in Roanoke, Virginia. I drank the Kool-Aid every day. I loved the bank, the team, and the impact we had working alongside small businesses in our community. We competed for loans against the biggest banks in the country, which meant we ran lean. We repaid TARP organically after the Great Recession, put up four straight years of record earnings, and ultimately sold the bank in 2015.
That sale forced me to figure out a next chapter. I started a consulting practice focused on community banking, and over four and a half years we grew to a team of 10 consultants working with financial institutions across the community bank spectrum. Every tech stack, every level of sophistication. We ran finance and accounting, strategic planning, and process improvement, and built an M&A niche that sat inside 13 transactions.
What we learned in those four and a half years is what made KlariVis inevitable. The same data struggles I had lived through as a CFO existed in every single bank we walked into. It did not matter how sophisticated the bank was, what core they ran on, or what their tech stack looked like. Everyone was buried in manual spreadsheets and living off stagnant reports. As CFO I used to get 15-plus emails a day, each with a different “report,” and it was my job to dig through them looking for the nugget that mattered. Most days I never got to half of them.
We knew there had to be a better way, and that is what drove us to build KlariVis, the first data intelligence platform built from a community banker’s perspective. Our consulting clients rolled their sleeves up and helped us build the original product. The goal was simple: empower bankers to get back to the business of banking, instead of spending days and weeks manually building reports that nobody reads because they are outdated by the time they land in an inbox.
That same foundation, the one that gets bankers out of the spreadsheet trap and into real decisions, is exactly what every conversation about AI in this industry now depends on. We will come back to that.
Q2: What foundational data moves should community banks be making today and why?
Let me start with what I’m hearing every day. The hype around AI in our industry is unreal. Every conference, every vendor pitch, every board meeting has someone talking about the shiny new AI tool that is going to change everything. I am a believer. We use Claude, ChatGPT, and other AI tools at KlariVis every day, and the productivity gains are real.
But here is the truth nobody wants to say out loud: you have to do the hard work first. AI accelerates institutions that already have a data foundation. It does not create one. You cannot ask any AI tool “what are my deposit trends” if your data is not structured, normalized, and accessible.
So, the foundational moves come down to three things.
First, get your data out of the silos and into a banking-specific data model. Your core, your loan origination system, your digital platform, your budget tool, none of them were designed to talk to each other. And banker to banker, this is not the place to go it alone. I have watched institutions spend millions and multiple years trying to build that model in-house, only to end up with a fragile system that one or two people understand. Your competitive advantage is not in building a data model from scratch. It is in how your bankers use the insights it produces. Partner for the foundation. Differentiate on what you do with it.
Second, get the governance right from day one. Customer-level loan, deposit, and transaction data cannot go into a consumer AI tool. It is a hard regulatory line. Any foundation you put in place has to handle personally identifiable information (PII) and regulated data appropriately, or you are going to learn an expensive lesson.
Third, start now. Too many banks treat this as a two- or three-year project they will get to after the next core conversion or strategic plan cycle. That is the wrong mental model. You do not need to boil the ocean. Get the high-value data elements working for your bankers in the next sixty to ninety days, and build out from there. That is exactly how we approach it at KlariVis. Our average bank is live on the data that drives deposit, loan, and margin decisions in under two months, with the broader foundation expanding over time. You do not have to wait for a multi-year data warehouse build to start getting value out of your data. Speed to value matters, and the institutions that wait until everything is perfect are losing ground every quarter.
Every bank that wants to use AI, which is happening right now, not in some future planning cycle, is going to discover that the AI piece is actually the cheap part. The expensive part is the ingestion, the normalization, the banking-specific model, the governance, and the institutional knowledge of what a banker actually needs to see every day. None of that gets cheaper because AI exists. If anything, AI raises the bar on how clean your foundation has to be.
That is the work. And it has to happen before the shiny AI tool delivers anything other than a confident-sounding wrong answer.
Q3: What changes can banks expect when lenders, branch managers, and other staff have access to clear, real-time data?
It changes the game entirely. When everyone in the bank is looking at the same data, in the same place, and can quickly spot the trend or the number that matters, decisions get faster and more confident. Gone are the days of wasting half a week trying to pull information together to answer a single question. It is right there in front of you.
One of our clients, a community bank CEO, told us that before the platform was in place, her team was running the bank off post-mortem reports. After implementation, her loan officers could see why a deal was off and how to fix it. Her branch managers came in the next morning seeing the impact of the prior day’s decisions, instead of waiting until month-end. In one case, a mispriced loan was flagged immediately. Instead of surfacing weeks later in a month-end report, the team had context right away. Was it a risk issue? Relationship depth? They coached the officer in the moment, fixed the deal, and the next deal that lender priced was better because of it. That bank ended up with a 50 basis point lift in commercial weighted average loan yields in the first year, and moved from a one-size-fits-all deposit pricing model to relationship-based retention.
That is what daily, accessible data does. It gets the friction out of the system. It gets the right information in front of the right person at the moment the decision needs to be made. And it builds a culture where people are acting on insight instead of waiting on a report.
This is also where the AI conversation gets interesting. The institutions that get to this state, where every banker has the data they need at their fingertips, are the same institutions that will be ready to layer AI on top of it. The ones still living in spreadsheets and stale month-end reports will be trying to bolt AI onto a broken foundation. It will not work.
Q4: In a similar vein, what does “usable” or actionable data look like in practice for someone making day-to-day decisions?
This is the ultimate goal. Getting the right data into the hands of the people actually making the day-to-day decisions that drive performance, profitability, and the level of customer service community bank customers depend on us for.
Let me make it concrete. One of our clients called us recently to share a story. A relationship banker spotted an unexpected $250,000 deposit hitting a customer’s account. Not a withdrawal, a deposit. The banker picked up the phone, thanked the customer, and asked how the bank could help put that money to work earning more for them. That conversation ended with the customer moving another $750,000 to the bank.
That is community banking at its best. And it almost never happens at a bank running on month-end reports. The data has to be in front of the banker the next morning, on a screen they actually look at, in a format that flags what matters. If the banker has to call finance to find out which of their relationships had unusual activity yesterday, the call never gets made. The deposit sits, the moment passes, and at some point that customer takes the larger sum to a wealth manager somewhere else.
Usable data is easier to define by what it is not. It is not a 40-page PDF that lands in your inbox Monday morning. It is not a dashboard nobody opens because it answers questions nobody asked. It is not a report that requires a specialist to translate. Usable data is role-specific, in the flow of work, and answers the question “what should I do today?” rather than “what happened last quarter?” It puts the banker back in the position of being the trusted advisor their customer expects them to be.
Q5: What are the most common gaps or flaws you see in how community banks approach data strategy and reporting today?
The biggest one, by a wide margin: most banks have no idea what the status quo is actually costing them.
We have done studies with our clients, before and after implementation, to quantify the hours their teams spend pulling information together. Recurring board packets, committee meetings, sales meetings, ALCO, exam prep. But also the ad hoc grind that nobody sees on a P&L. The special request that comes in on a Tuesday afternoon. The same question being asked of three different people in three different departments, each one writing their own version of the same report. The back-and-forth between the person who knows how to write the report and the business user, where the data writer cannot understand why the dashboard they built is not being used, and the business user cannot understand why they cannot get the information they need to act. That cycle is endless, and it is everywhere.
When we add it all up, the average two-billion-dollar bank is spending over 50 thousand dollars a month on the status quo. Every month. And that number does not include the opportunity cost of what those people could have been doing instead, the cost of decisions that got delayed because the report was not ready, or the key-person risk that comes with one or two people in finance holding all the tribal knowledge of how the spreadsheets fit together.
One of our clients, a CFO at a community bank just over one billion in assets, put a number on the other side of that equation. After moving onto KlariVis, her bank expanded net interest margin by almost 40 basis points in the first year, because the team finally had daily visibility into the deposit and loan decisions that drive it. That lift did not come from a new strategy. It came from bankers across the institution making better decisions in the moment, because the data was there.
We cannot keep doing it the way we have always done it. AI is changing that whether community banks are ready or not. And the institutions that are still spending 50 thousand dollars a month rebuilding the same spreadsheets are going to find themselves on the wrong side of that shift very quickly.
Q6: Some community banks consider building their own data platform. What are the biggest misconceptions about what it takes to do that successfully?
The misconception I see most often: “we already have a BI tool” or “AI will let us build our own.” Both underestimate, by an order of magnitude, what the work actually is.
Building a banking-specific data foundation is exactly that. Hard work that takes time. You have to get the data out of your core, your loan origination system, your digital platform, and every other system it lives in. You have to normalize it. You have to design a data model that reflects how a bank actually works, household, relationship, product, risk. You have to stand up governance to handle PII and regulated data. And you have to have someone on the team who can translate banking concepts into technical structure, because bankers are generally not data engineers, and data engineers generally do not understand banking.
And here is the part bankers rarely think through before they decide to build: a data warehouse is not a project. It is a permanent expense. Every time you add a new product, the data model has to be updated. Every time a vendor changes their file format, the ingestion has to be rebuilt. Every new regulatory or reporting requirement lands back on the same internal team. A core conversion, when it does come around, can blow up your extracts and force a meaningful portion of the build to start over. The data engineers and database administrators you hire to build it have to be retained to maintain it, and good ones are expensive and hard to keep. The build itself takes years. The maintenance takes forever. That is the part nobody puts in the original business case.
The misconception is that any of that is fast, or cheap, or a one-time investment.
While the IT team is heads-down on the build, the people who are actually supposed to be moving the needle for the bank, the lenders, the branch managers, the CFO, are still working off manual spreadsheets and static reports. The status quo cost we talked about earlier keeps running. Every month.
A client told me recently that they embarked on a data warehouse project with a vendor to stand up a major cloud data warehouse, in parallel with bringing us on. About six months into the warehouse project, they realized what it was actually going to take. They needed to hire a database administrator. They needed to build data pipelines. They needed QA resources. They had to figure out how to get data out of their siloed systems. He looked at me and said, “this is not what we do. We are bankers. We have no business doing this.” And they halted the project entirely. Meanwhile, his banking team is leveraging KlariVis every day to make the decisions that matter.
That is the trade-off that nobody is being honest about. You can spend the next two or three years building the foundation yourself, with people you have to hire and a roadmap that keeps expanding. Or you can partner with a provider whose whole company is built around solving this problem for community banks, and have your bankers operating on a real foundation in sixty days.
AI does not change this calculus. If anything, it makes the build path harder, because the AI piece is the cheap part. The expensive parts, the ingestion, the normalization, the banking model, the governance, are exactly the parts AI cannot do for you.
Q7: How should banks think about the trade-off between building in-house and partnering with a provider?
I think the better question, once you have decided to partner, is what to look for in the partner. Because not every provider is built for community banking, and the wrong choice can be almost as expensive as building it yourself.
A few criteria I would put at the top of the list.
First, was it built by bankers, for bankers? The banking-specific data model is the hardest part of this whole equation. A provider who came at this from a generic analytics background, or who serves bigger institutions and is trying to move down-market, will not have the model right. You will spend the next two years explaining your business to them.
Second, how fast does it deliver value? If the implementation timeline is 12 or 18 months, you are partnering with a build project, not a platform. The whole point of partnering is that your bankers should be making better decisions in months, not years.
Third, how does it handle your data and governance? Customer-level loan, deposit, and transaction data has to be handled the right way from day one. Ask hard questions about where the data sits, who has access, how it is protected, and what happens if you ever want to leave. Any provider who gets defensive about those questions should be off the list.
Fourth, how does it handle change? Cores get converted. Products get added. Regulations change. The right partner makes those transitions easier, not harder. The wrong one becomes another vendor you have to manage through every core conversion for the rest of the relationship.
Fifth, does it understand that this is the foundation for what comes next, including AI? The platform you partner with today should be designed to be the data layer your AI tools sit on top of tomorrow. If your provider does not have a clear point of view on that, they are not thinking about the future you are actually walking into.
The build-versus-partner decision is really the build-versus-the-right-partner decision. Partner for the data foundation with someone who understands this business. Differentiate yourself on how you leverage the data and the insights it produces. That is where your competitive advantage actually lives. And find a partner that is built for the journey, not just the destination. Every bank is at a different stage. The right partner meets you where you are and grows with you.
Q8: What are the hidden costs, beyond budget, that leadership should be considering in that decision?
The hidden costs are not on any budget line, which is exactly why they are dangerous.
Four worth naming.
Strategic blind spots. The decisions you do not make because the data is not there. The deposit you do not retain because nobody saw the balance trending down. The cross-sell that goes to a competitor because nobody saw the customer take out a loan somewhere else, or move money out for a down payment, or open a business account across the street. The unprofitable relationship you keep investing in because nobody surfaced the math. None of those losses show up on a P&L as “data foundation cost,” but they are real money.
Cultural cost. When leadership does not trust the numbers, decision-making slows across the entire organization. People hedge. They ask for one more cut of the analysis. They wait for the next meeting. That hesitation has a real cost, especially in a market where deposit attrition and margin compression do not wait.
Regulatory and exam cost. Examiners are asking harder questions about data governance every cycle. Can you reproduce a report two years later? Can you explain how a number was calculated and who touched the data along the way? Most banks running on spreadsheets and emailed reports cannot answer those questions cleanly, and the regulatory bar is rising, not falling.
AI readiness cost. Maybe the most expensive of all. Every quarter spent on the wrong path, building it yourself or postponing the foundation work entirely, is a quarter the institutions that got this right are pulling further ahead. By the time you arrive at the AI conversation with a broken foundation, the gap is no longer something you can close in a budget cycle.
Q9: How does the strength of a bank’s data foundation impact its ability to adopt more advanced capabilities like AI?
Let me be as direct as I can: AI is the reason every community bank needs a strong data foundation. There is no AI strategy without a data strategy. It is not a reason to wait on building one, and it is not a substitute for the foundation work itself.
Here is the mechanics of it. Say you want to use AI to predict deposit attrition. A reasonable, high-value, very banker-relevant use case. To produce a useful answer, an AI model needs household-level deposit balances tied to the right customer, transaction history connected to that same customer, product mix, tenure, channel behavior, and ideally some signal about the broader relationship like loans, debit card activity, and digital engagement. All of it normalized. All of it accessible. All of it tied to the same customer identifier across systems.
Most community banks cannot produce that today. Not because the data does not exist, but because it lives in five different systems with three different customer IDs and no connective tissue between them. You can buy the most sophisticated AI model on the market and point it at that environment, and it will produce a confident-sounding answer that no banker should act on.
This is the part that gets lost in the AI hype. The model is the easy part now. The foundation is the hard part. And customer-level loan, deposit, and transaction data cannot go into a consumer AI tool like Claude.ai or ChatGPT. So “we will just use AI with our bank data” is not actually a plan. It is a regulatory problem waiting to happen. The serious version of using AI with bank data requires a governed, structured data foundation inside a purpose-built environment that was designed for it from day one. Bank data is sensitive, regulated, and complex. It belongs somewhere built to handle it. That is exactly what the foundation work is for.
This is exactly the work we have been doing at KlariVis for years, and it is exactly the work the AI moment is now demanding of every community bank. The foundation we have built for our clients, normalized banking data, governance built for regulated information, household and relationship-level views of the customer, is the same foundation any serious AI strategy will require. Banks who are already on the platform are not behind on AI. They are ahead, because the hard part is done.
Most banks have the AI narrative inverted. They think the foundation is the unglamorous prerequisite to the exciting AI work. The reality is that the foundation is what creates the AI opportunity in the first place. Without it, AI produces fluent, confident, wrong answers. With it, AI starts giving your bankers genuine leverage.
Q10: What are the biggest risks of trying to implement AI before the data foundation is ready?
Three risks, in order of how badly they will hurt you.
The first is wrong answers, delivered confidently. AI models do not hedge. They produce fluent, plausible, authoritative-sounding output whether the underlying data is clean or not. Anyone who uses these tools regularly has seen this firsthand. Two people on the same team can give Claude or ChatGPT the same prompt and get two different answers. That is manageable when you are drafting an email or summarizing a document. It is not manageable when the answer is going to drive a deposit pricing decision or a credit call. If your data lives in three systems with three different customer IDs, the AI tool will still answer the question. It will just be wrong. And your team will act on it, because that is what tools are for.
The second is regulatory exposure. Examiners are going to ask how a decision was made, what data informed it, who governed that data, and whether you can reproduce the answer. If your foundation is broken, you cannot answer those questions cleanly. The institutions running AI on top of unstructured, ungoverned data are building a regulatory problem that will surface on a future exam cycle.
The third, and the most damaging long-term, is erosion of trust. Once the board or the executive team sees AI-driven output that is wrong, adoption stalls for years. People go back to the spreadsheets. They build workarounds. You do not get a second first impression with AI inside a community bank.
I want to be precise about something I am hearing from banks right now. Plenty of institutions are bringing in outside firms to automate a narrow piece of underwriting, or fraud, or document handling. In a bounded use case like that, where the data is relatively self-contained, those projects can absolutely deliver value, and I am not going to tell anyone they are wrong to be moving. What I would gently raise is the bigger picture. Each one of those point projects, by itself, becomes another silo. The underwriting model does not know the depositor relationship. The fraud tool does not see household profitability. If you run three or four of these with three or four different firms, you have not solved fragmentation. You have made it more expensive. The foundation is what keeps those projects from compounding into the next version of the problem you are trying to solve.
None of this is an argument against AI. It is an argument for sequencing and connective tissue. Foundation first. Then AI. In that order, every time.
Q11: Looking ahead, what will separate community banks that successfully turn data and AI into a competitive advantage from those that don’t?
It will not be technology. It will be discipline.
The winners will be the institutions that built the foundation while everyone else was still talking about AI. They will be the institutions whose lenders walked into Monday morning meetings knowing exactly which relationships needed attention, while their competitors were still waiting on a month-end report. They will be the institutions whose CFOs were managing margin in real decisions, not in retrospective explanations. They will be the institutions whose CEOs treated data as a strategic asset with executive ownership, not as an IT project to be delegated.
And when AI gets layered on top of that foundation, those banks will pull away. Not because the AI tool is different. They will be using the same models everyone else is using. They will pull away because the foundation underneath the AI is real, the data is clean, the governance is in place, and the bankers know how to use it. And there is a second-order effect that is easy to miss. The banks that built the foundation will not just be running AI. They will be running it on connected data, so every model they stand up, whether it is underwriting, deposit attrition, fraud, or the next use case nobody has thought of yet, draws on the same governed, relationship-level view of the customer. The banks that went project by project, firm by firm, will be running a collection of disconnected tools that each see a sliver of the customer. Same models. Completely different leverage. The institutions still running on spreadsheets will be trying to bolt AI onto a broken foundation, and they will spend the next several years either falling behind or rebuilding from scratch.
Here is what I would say to any community banker reading this. The competitive advantage of a community bank has always been the same. You know your customers. You know your market. You make decisions faster and with more context than the big banks can. Data and AI do not change that advantage. They compound it, if you build the foundation now. And they erase it, if you wait.
That is the choice in front of us. And the institutions that decide right are going to be writing the next chapter of community banking.