Starting a Bank in the Age of AI

Coming up, we're back in the studio. It's me, one of your hosts, Al Dominick, with my friend Steve Williams back at headquarters in Scottsdale. Good to see you, buddy.

Good to be back at HQ.

Today we're going to do something a little different from previous episodes. We've had some great bank CEOs share how they're trying to position themselves for future growth and success. I thought we'd pause and come up with some prescriptive ideas for how we would build a bank ourselves if somebody gave us the opportunity.

There's a lot happening, and so much of it is AI-driven and technology-focused. If Steve and I take a few minutes to get Plugged In around where and how we might build a really valuable franchise, maybe it will be useful for listeners.

I heard a stat recently, and I can't remember the name of the company, but for the first time in the AI era, we've apparently seen a one-person company create a $1 billion valuation.

Think about all the great financial institutions that worked for 75 years to create a billion-dollar market cap, and now one person can potentially create that kind of value.

That's obviously not the norm, but it illustrates the shift in how value is being created. You know the song Mad World? It feels a little like that right now.

Every couple of decades there's a major wake-up and adjustment. You can feel that this year.

What's funny is, as you're singing, I took some liberties in advance of this because if we're talking AI, we should talk about some of the tools we're both experimenting with.

I've been a little more open on LinkedIn about exploring vibe coding. There's a tool called Kiro that some fintechs are looking at as they think about their technology stack and code quality. I figured if I could start dabbling in this myself, that would be useful.

You and I have also used Suno for music. I thought Plugged In could use some new theme music, so I made you something.

Okay.

You've got Nashville. I want to see some action.

All right. I could play it all day.

You've got a Nashville-style country version of Plugged In.

We had to have a little fun. But that little experiment shows how creative you can be in a very short period of time.

That's part of why I wanted this episode to focus on a simple question: if I were building a bank from scratch today, what would I think about, and what would I leave behind?

People like us can scratch that creative itch now. If I want to create a song, I can do it without a band or a studio. I can produce something in minutes that might actually be useful or entertaining.

Businesses are facing a similar shift. We have clients trying to figure out where they should participate in newer technologies, how those tools fit the outcomes they're pursuing, and how they work alongside the delivery channels and operating models they already have.

So let's reverse-engineer it. In this conversation, we bring no technology debt and no legacy mindset. We're not heavily scripted. We've both been traveling a lot, so we've got anecdotes we can throw in.

If I were starting a bank today, I'd first look at you and ask: what are we actually good at, and what can we do that will materially matter to our potential customers?

A recent example is Nubank, which started in Brazil and has expanded into markets such as Mexico and Colombia while moving toward the U.S. They passed the 100 million-customer milestone, one of the first digital banking companies to reach that scale.

What I find interesting is how leadership describes the next strategic shift. They believe they succeeded as a digital-native bank and now want to become an AI-native bank. They were early in acquiring Hyperplane to strengthen that capability.

A lot of traditional banks are still working toward digital maturity, so it's a useful contrast.

Investors are already pressure-testing AI disruption across industries. Whether you look at the recent volatility in software-company valuations or private equity firms assessing every portfolio company for AI risk, they're asking what this means for enterprise value.

Banking needs the same kind of stress test.

One hundred percent. Imagine a well-funded, AI-native company coming in and saying, "I'm going after your top 20% of customers, the most lucrative, supposedly loyal segment you have."

If it creates a compelling offer, what keeps those customers from at least taking a look? And if they like what they see, what prevents them from moving?

That's the uncomfortable question.

There are a lot of tools worth experimenting with right now. Kiro is one. Suno is another. You have tools for coding, agents and visual creation, including things like Nano Banana. Some of the names sound ridiculous when you say them out loud, but these are valuable learning opportunities.

If I were building something from scratch, I'd be using tools like these because they can help me get smarter faster about discrete opportunities where we might win.

We've talked about finding riches in the niches, and AI can help identify those opportunities.

Jeff Bezos famously said, "Your margin is my opportunity." If I were going after a top-tier commercial bank as a startup, one advantage I'd want from day one is a clean data model.

I'd start by defining the commercial businesses we want to serve, how we manage those relationships, how credit information is structured and what information we need to support everything built on top of it.

A lot of established banks are still cleaning data simply to create a usable foundation for AI. A startup has the advantage of designing that foundation correctly from the beginning.

I can tell you've been a consultant for 20-some years because every consultant starts with data.

But you're right. You need a strong framework around who you want to serve, how you're going to differentiate and what you're bringing to those customers.

Then technology can help identify patterns very quickly. It gives you the ability to move, but more importantly, it can show you where a niche opportunity exists that you might otherwise miss.

That's how I'd think about building the bank. I wouldn't spend all my time worrying about the charter. You can buy or obtain a charter.

I'd rather ask whether we can assemble a small tiger team of highly curious people who understand business applications and can build something valuable. I'd bring in people who understand the regulatory and compliance realities of banking, but I'd keep the team tight.

If we start with strong data and give a small team the right tools, we may identify opportunities that a traditional organization would never have considered.

Think about the information a bank naturally collects. We learn a lot about customers because we lend to them. We also facilitate their payments and deposits.

Put those information sets together, and you can build substantial intelligence. Every day you can monitor commercial clients for credit risk and payments activity, but you may also be able to give useful intelligence back to them through digital channels.

If I can personalize that experience and deploy agents through a treasury platform to help commercial customers accomplish things, I'm adding value as a trusted adviser rather than merely providing transactions.

You brought up agents. I was in Nashville recently, which is where the country-music inspiration came from, and I talked with several executives running banks between roughly $2 billion and $5 billion in assets.

They're all curious about agentic commerce and what is happening, but they don't know where to start. They don't necessarily know how to build agents or who is creating agents for banking use cases.

If you and I were starting a bank together, I'd build the brand and operating expectation around speed. I would incorporate agents into the process from day one rather than taking old workflows and trying to rearchitect them later.

That aligns with what we're seeing in the broader regulatory environment. The administration has been friendly to innovation, the GENIUS Act has accelerated stablecoin conversations, and suddenly everybody seems to want some kind of bank charter, trust charter or related authority.

They're applying, acquiring or looking at trust powers.

One theme is becoming clear: technologists are absorbing banking knowledge faster than bankers are absorbing technology knowledge.

That's exactly why, starting from scratch, I'd build a tiger team of technologists who aren't constrained by, "This is how I grew up in banking."

I'd give them the problems to solve, then bring in KYC and compliance expertise and somebody with deep credit judgment. The fresh technology perspective comes first, with experienced banking knowledge layered in to make sure the solution works in the real world.

You don't ignore history's lessons. It still comes back to people, products and performance.

Think about our conversation with Mike Daniels at Nicolet in Wisconsin. He described a culture where the team could lock arms, understand exactly how they wanted to serve their market and consistently compete for the upper end of the customers they wanted.

I'd recruit people with that same desire to win. They don't compete with each other internally. They understand that each person brings different skills to the story, and they recognize that AI-enabled tools need to become fundamental to how the organization works.

Security obviously has to be part of that. Tools like Claude are becoming incredibly powerful, and the same capabilities that make them useful can create new risk.

If I were a banker today, I'd make sure my CISO and technology team could create a safe environment, isolated appropriately from the rest of the enterprise, where smart people can experiment with agents.

At Cornerstone, we've held an AI Day for several years where we ask people to show what they're doing with AI. Recently, three team members demonstrated agents they had built themselves. One of my favorite details was that the winner of the competition was the oldest person in the group.

You need a safe place for hands-on experimentation. There are probably people inside every bank already experimenting at home or wishing they could do more at work. Give them clear guardrails and a secure sandbox while taking infrastructure, privacy and data security seriously.

If I were starting a bank, I'd also be recruiting smart younger people who live in this technology every day in a way that's different from you and me after 25 or 30 years in business.

We can help teach them banking. They can help teach us how new digital tools can be manipulated in clever and creative ways.

Small teams moving with speed and purpose are important. The goal isn't to break things randomly. It's to make experimentation part of the operating culture.

I was struck recently by a bank CEO's experiment with OpenAI during an earnings call. For roughly the first 20 minutes, an AI-generated version of the executive delivered the results, and the company used it to introduce an OpenAI business relationship it was proud of.

That's a roughly $20 billion-plus bank demonstrating that this technology is moving fast and that a traditional financial institution can experiment publicly.

I had initially thought Jay had done it, but Sam was the one who reminded us of the example. Either way, it was a clever splash moment.

The executive had once talked with me about how the NBA creates these big moments that capture attention, and how he wished banks would occasionally do something similarly memorable. This was one of those moments.

If we were building a new bank, I don't know that I'd want an agent representing the leadership team on an earnings call, but I absolutely want that kind of creativity inside the company.

Then we have to connect the experimentation to performance. What are we going to measure? What creates value over time, and how can newer tools help us uncover it?

The efficiency ratio will obviously matter, but I also like revenue per employee.

An average bank might generate roughly $250,000 to $300,000 of revenue per employee. A digital innovator such as Chime can be dramatically higher.

That creates a fascinating question: what revenue per employee could a bank achieve if intellectual property, automation and agentic work were embedded throughout the operating model?

For a commercial bank, I'd also think about a concept we're hearing more often: service as software.

You keep the human relationship at the last mile, but almost everything behind that relationship becomes software-driven.

Banking has historically taken great pride in relationship managers and credit officers. That's fine. But the work supporting those people should become increasingly intelligent and automated.

The expense structure then becomes a high-value human layer supported by software, with a smaller number of very capable people managing models, agents and automation.

Agile teams with human relationships in front are a compelling model for the future.

That ties into the broader movement-of-money themes we've discussed on previous episodes. Stablecoins may have upside, especially in cross-border transactions.

Depending on the bank we created, we'd have to decide where and how to participate in tokenized deposits or stablecoins. I don't see a small startup bank issuing its own stablecoin, but community banks clearly want to understand where they fit in these conversations.

They won't make good decisions unless they're clear about what they want to do and what they don't want to do.

If a bank says, "We're not wasting time over here, but we do have international or cross-border customers who could benefit from this," that's a much more credible starting point.

I've been in meetings where AI, blockchain and stablecoins all collide around one central banking issue: cost of funds.

Banks have historically benefited from funding that is cheaper than the brutal daily market rate for money. People have predicted for decades that better information would eliminate idle or underpriced deposits.

It didn't happen when interstate banking expanded. It didn't happen with the internet. Twenty-five years ago, analysts predicted that all "dumb money" would become smart money. Yet today, some banks still have extraordinarily low funding costs.

Relationships are one reason. Customers leave money where it's convenient and where they see value.

Stablecoins and AI may intensify that competition for funding more than mobile banking, digital banking or interstate branching ever did.

But they also create an opportunity. Banks can add more value to the movement of money through smart contracts, specialized services and intelligence they give back to customers. If customers see that value, they're more likely to keep money in the ecosystem.

This cost-of-funds issue concerns people in a way I haven't seen in decades.

We could keep designing this hypothetical bank, but I'm not sure starting one is where Steve and I should actually spend our time and talent.

The intellectual exercise is useful, though. If you remove the legacy guardrails and ask what you would build today, it forces you to think differently.

An established bank doesn't have that luxury. It has regulatory expectations, compliance responsibilities, privacy obligations and existing infrastructure. But technology is moving too quickly to ignore the question of where the institution could operate differently.

So I brought a few notes for an established bank. Imagine a $10 billion to $20 billion institution that isn't starting over but wants to decide what to protect, what to fix and what to let go.

I'd start by defending the core business. Then I'd move to the edges and identify where the bank is slow, expensive or simply doing things that no longer make sense.

You've encouraged me before to think about a kill switch for investments that don't produce value. That return-on-technology concept becomes even more important as AI changes the economics.

Ron Shevlin once asked me whether technology can still be a competitive advantage. My answer was no, not by itself. The advantage comes from aligning technology with a distinctive business model, niche and strategy.

The winners allocate both capital and operating expense toward something that passes the Michael Porter test: they do things differently from competitors in ways customers genuinely value.

Then they use new tools to reinforce that advantage.

I also think about the leadership team of the future. It can't be composed only of people with wisdom who oversee hierarchies.

Future bank leaders will increasingly design and deploy systems and models. They'll understand what technology can do, collaborate effectively with technologists and build systems that perform work.

That's one of the biggest differences between a fintech workforce and a traditional bank workforce.

Those people will also view data as an asset rather than a cleanup project. They can use pattern recognition more thoughtfully, uncover things others may overlook and allocate capital more intelligently.

Anecdotally, I was recently with Kim Snyder, who founded KlariVis. She was talking about the early days of the company. A handful of banks trusted her enough to become early customers, and she was able to identify very specific problems CFOs needed solved around the data they craved.

Banks should take that same mindset internally. You don't necessarily need to start a new company. Ask what discrete problems remain unsolved and whether data you already possess can bridge the gap.

Banks are sitting on a treasure trove of information. The challenge is unlocking it.

For larger institutions trying to reconsider what's possible, I would go much deeper with data and make sure I understand the regulatory advantages banks still possess.

At the same time, pay attention to legislation and policy changes that are encouraging new entrants. In some ways, it feels like an Oklahoma land run into banking right now because of what's happening with AI, blockchain and stablecoins.

That is another reason banks shouldn't limit their recruiting the way they sometimes do.

If I were a smart, talented person looking for an industry going through a fascinating period of change, I could go work for Chime or Stripe. But I could also join an established bank that is trying to reposition itself and already has brand equity, customer loyalty and capital to deploy.

That could be a very rewarding career.

Pair those people with experienced bankers who understand credit, compliance and loan servicing. Let a tiger team push on the old assumptions and ask where waste can be removed or where the process should be redesigned completely.

Two AI opportunities stand out. Banks have enormous amounts of unstructured data that can now be analyzed far more quickly. Think about every credit memo sitting somewhere in the organization. You can use modern tools to read those documents and surface trends that were previously hard to identify.

The same thing applies to conversations. Call-center interactions or senior loan committee discussions can become a source of continuous insight rather than disappearing after the meeting.

Those are pragmatic ways a bank can start changing how it operates this week.

At the individual level, we also have to walk the talk. There comes a point where leaders need to allocate time to get uncomfortable with technologies that will affect their business.

I mentioned vibe coding because I've been in conversations where people get excited about it and I realize I don't have as deep a knowledge base as I want. The only way to close that gap is to try it myself.

Touch the clay.

Exactly. It may be ugly at first. It may be a little embarrassing, but touch the clay.

Curiosity is a leadership attribute, and this is where leaders have to model it.

If a younger person sees me make something useful with a tool and thinks, "If Al can do that, I can do it faster," great. Then we learn from each other.

This is a remarkable moment for bank executives. There is a legacy worth protecting, and there are parts of the business that need to grow and change in ways that may feel uncomfortable.

The answer isn't command and control from a few big-company partnerships. Everybody needs to get their fingers dirty with this technology.

You learn by doing. And once you do, you realize this isn't the old world of plan, design, build and deliver in a long sequence.

Now it's plan, design, build, deliver, learn, then do it again. The cycle keeps compressing.

That's why, if we're going to be in the studio together, I wanted the conversation to be conceptual and practical at the same time.

We've got some great entrepreneurs joining us in upcoming episodes. I love hearing how they've built things and how they refuse to become satisfied with the status quo.

Past guests such as Mike Daniels and Jay Hillebrand from Stock Yards have talked about building on a strong legacy without becoming a sleepy institution that loses its edge.

Continuing to stay curious and creative is a big part of why we get Plugged In.

Thanks so much, buddy. Good to see you.

Good to see you.