Transcript
Hey everybody. Welcome back to another episode of What’s Going On in Banking. I’m Ron Shevlin, chief research officer at Cornerstone Advisors, here with my co-host Stacey Bryant.
Stacey, what’s going on?
How are you doing, Ron?
I’m doing fantastic.
I just got back from the North Dakota Bankers Association Bank Management Conference, which was held in Arizona. You and I have spent what feels like the entire months of January and February in Arizona, which is great for those of us escaping the Northeast winter.
Shout-out to Dorothy Lick and everybody at the association for having me.
They asked me to speak about AI, what banks are actually doing with it, what community and regional banks should be thinking about, and where data strategy fits.
I also gave out an unofficial Stacey GonzoBanker award to somebody in the room: Don Morgan from the Bank of North Dakota.
The Bank of North Dakota is working with Fiserv on a U.S.-dollar-backed stablecoin called Roughrider.
Great name.
They’ve also been talking with banks and credit unions around North Dakota and the broader Midwest about what stablecoins could mean for institutions and issuers.
So stay tuned. I’d like to get more of that conversation going because they’re doing something interesting in the retail-crypto and stablecoin space.
How are you doing?
I’m also coming back from Arizona.
I left four-degree weather in Boston for 80-degree weather, which was pretty great.
This was my fourth Arizona trip in a row.
My wife and I spent a few days around Tucson doing some hiking.
Did you get any personal time while you were out there?
I did.
After my speaking session, I took the kids to Sedona. We did a beautiful hike through the red rocks and tried unsuccessfully to get a Christmas-card picture.
Then we drove to the Grand Canyon.
And the Grand Canyon is exactly what it says it is, Ron: grand.
It was incredible.
The only “mountains” I grew up with were New York City skyscrapers, so seeing that kind of landscape always makes me stop and reflect.
While both of us were out hiking in different parts of Arizona, a lot was happening in banking.
Our last episode ended with you saying, “Stacey, next time we have to talk about Capital One and Brex.”
So let’s do it.
This news is a couple of weeks old now, but it is still important.
Capital One announced a roughly $5 billion deal for Brex.
That is a substantial price, although lower than some of Brex’s earlier private-market valuations.
A lot of fintech people are understandably excited because it is a major exit.
But I think the more useful question is whether the combination actually works.
Overall, I’m positive on it, but there are risks.
The first is culture.
Can a startup like Brex thrive inside a large bank?
Capital One is not a traditional bank in many respects, so maybe the answer is yes, but integration still matters.
There’s also a customer question.
Why do small and midsize businesses use Brex?
Technology is a major reason, but some customers may also like that Brex is not a traditional big bank.
Once those customers become Capital One customers, does that change the relationship in a way they don’t like?
Then there’s basic integration capacity.
Capital One is still digesting Discover. Adding Brex on top of that is a lot to absorb.
That said, I have a lot of respect for Capital One’s strategic discipline.
They generally have a clear idea of what they are trying to build.
So I think the deal can work, but there will be real hurdles over the next couple of years before they fully realize the value of a $5 billion acquisition.
From a former banker’s perspective, two things jump out at me.
First, small and midsize business customers are increasingly being trained to expect integrated cards, accounts payable, accounts receivable, expense management, and eventually newer rails such as stablecoins in one experience.
A bank cannot assume that “we have online banking and a good treasury-management rep” will remain enough.
That proposition could age like MySpace.
Remember MySpace and ranking your top eight friends? Thank God we don’t have to do that anymore.
Second, if you are a fintech with a horizontal spend-management or card-plus-software product, the competitive bar just moved.
You are not only competing with other startups anymore. You are competing with bank-fintech combinations backed by enormous balance sheets and technology budgets.
That changes the exit market too.
If you are building a fintech, the goal becomes creating something so vertically integrated and mission-critical that a large bank sees you as a time machine, not simply a feature.
Capital One gets Brex’s technology, product velocity, and a serious position in business payments and software.
Brex gets scale, capital, regulatory infrastructure, and the ability to move faster on parts of its roadmap.
The broader message to the industry is that the next phase of B2B banking may involve banks owning software, not only sponsoring it.
And if you run a community bank or fintech, you are no longer competing only on rates and relationships. You are competing for ownership of the workflow.
I think there is also a segmentation lesson.
For small and midsize businesses, I see at least three major dimensions that shape banking relationships.
One is access to capital. For many small businesses, the most important question is still, “Will you lend me money when I need it?”
Second is industry specialization. Do you understand my business well enough to provide products, advice, and services that fit my sector?
Third is technology. Do you give me tools that make running the financial side of my business easier?
You could add a fourth dimension around human service, but I’m not sure that alone drives enough relationships anymore.
The Capital One-Brex deal is a clear bet on the segment that prioritizes technology.
There will be lending and card opportunities, but the reason many businesses chose Brex in the first place was the product experience.
For that segment, software is part of the product.
That creates a challenge for midsize banks. They need to decide which SMB segments they really want to serve and what the product strategy is for those segments.
This is a huge growth opportunity because many small businesses remain underserved.
We did research a couple of years ago showing that small businesses may borrow from community banks but often keep their primary operating accounts with the mega banks.
Capital One is positioning itself even more directly against Chase, Wells Fargo, and Bank of America in that space.
That is why this acquisition matters beyond the $5 billion headline.
And it brings us right back to a theme we constantly revisit: the riches are in the niches.
What segment are you actually trying to serve? What problem are you solving? What technology or partnership gets you there?
Don’t acquire or partner simply because everybody else is doing it.
Define the objective first.
That connects nicely to another article I saw about Goldman Sachs and Anthropic.
Goldman is essentially saying, “We don’t need every software vendor. In some cases, we can build our own AI workers.”
Goldman has been working directly with Anthropic, the company behind Claude, to build autonomous agents for areas such as trade accounting, transaction reconciliation, and client onboarding, including KYC and AML processes.
Anthropic engineers have reportedly spent months embedded with Goldman teams, working directly alongside bankers and technologists.
So this is more than a superficial partnership announcement.
Goldman’s CIO, Marco Argenti, has described these systems as digital coworkers for highly scaled, complex, process-intensive work.
The use cases target enormous pain points.
Trade and transaction reconciliation can take days. An agent may help flag or resolve issues much faster.
KYC and onboarding can take weeks across jurisdictions. AI-assisted document review and compliance checks could compress that cycle.
What interests me is the build-versus-buy implication.
If a large institution such as Goldman can work directly with an AI lab and potentially replace multiple software tools with an agent-based workflow, what does that mean for the software middle layer?
I have a couple of reactions.
First, those agents still have nothing on me. I don’t need PTO either, and I don’t complain about the coffee because I work from home and make my own.
More seriously, let’s look at this from the perspective of midsize banks and credit unions.
They are not going to have Anthropic engineers sitting inside their organization building custom agents from scratch.
For them, there are two more immediate challenges.
The first is clarity around what an AI agent actually is and what it does.
When I hear the Goldman story, some of the language makes it sound like agents are replacing whole enterprise applications.
That is not how I would define most agents today.
An agent is generally task-oriented. It completes a task or a series of related tasks.
It does not automatically replace the full ERP, accounting platform, loan-origination system, or every workflow around it.
The second challenge is not really build versus buy.
For most midsize institutions, the answer is buy.
The harder question is whether they buy specialized agents today from smaller vendors and integrate them into existing systems, or wait two or three years for major application providers to build or acquire those capabilities themselves.
Take commercial lending.
There are companies such as ENFi building agents that can integrate into loan-origination workflows and perform specific tasks.
That may create value now, but the bank still has to integrate, govern, and manage the new capability.
So is the juice worth the squeeze today?
Or do you wait for your core or LOS provider to incorporate it later?
And if agents eventually appear across hundreds of processes, where does prioritization come from?
What should the institution automate first?
I would be willing to bet that if you asked Goldman’s CEO a few years ago what the company’s first big AI-agent priority should be, “trade accounting reconciliation” might not have been the first answer.
That use case probably emerged from a functional team that understood the pain deeply.
For midsize institutions, we are heading into a very complicated prioritization challenge.
That reminds me of an article our colleague John Meyer wrote about the “invasion of the job snatchers.”
He made the point that even when AI automates work, somebody still has to monitor the models, retrain them, check for drift, and make sure the helpful banking assistant does not become a compliance or litigation problem.
Think about lending and all the bias risk that comes with it.
I’m going to tell one of John’s jokes because it actually gets at the point.
A guy walks into a grocery store on a Friday night with a six-pack of beer, hot dogs, and chips.
The cashier looks at him and says, “You must be single.”
He’s shocked and says, “I’m a data scientist. I spend hours analyzing information to make accurate predictions. How did you know?”
And she says, “Because you’re ugly.”
The moral, beyond the joke, is that the obvious human context can be very different from what a model assumes it needs to analyze.
We are not at a point where human judgment, emotional intelligence, and context disappear.
That is why midsize institutions should not panic when they see Goldman making investments they cannot replicate.
There is still enormous value in people, process knowledge, and targeted innovation.
I have no response other than I hope we don’t get complaints about that joke.
Our audience is very good-looking, so we should be safe.
All right, tell us about Array and Chimney.
There have been a number of high-profile fintech acquisitions recently.
MrBeast’s organization acquired a fintech. The Asian super app Grab acquired Stash in the U.S.
But there is another deal that I think is flying under the radar: Array acquired Chimney.
Array provides personal-financial-management and data-related tools. Chimney uses property data to help consumers understand and manage home equity.
I think the deal points to a bigger trend.
First, most financial-institution checking accounts are still fundamentally generic products. Banks try to differentiate them with messaging, features, and service, but the underlying product often looks very similar from one institution to another.
Second, many banks and credit unions talk about financial wellness, financial literacy, and financial health but do relatively little to actually change customer behavior.
Too often it is a library of PDFs and educational articles.
That is why I think there is an interesting competition emerging between companies such as Array and StrategyCorps.
StrategyCorps has been in this space for years with products that bundle third-party benefits into checking accounts.
They work with providers that offer things such as protection, discounts, and other services, and the goal is to differentiate the account while generating revenue.
Array is coming at it from a more digitally integrated direction.
Services such as data-breach protection, subscription management, credit tools, and now home-equity insights can sit inside the digital-banking experience.
Both models are trying to turn checking from a generic account into a bundle of useful capabilities.
And the important part is not only adding features. It is measuring whether those features deepen the relationship, generate revenue, improve cross-sell, or change behavior.
Almost every bank offers access to a credit score now.
Okay. What does that actually do?
Does it change the customer’s financial behavior? Does it lead to a loan relationship? Does it create engagement?
We have written about this with providers such as Bloom for credit-related services and Investi for investment capabilities.
I think Array and StrategyCorps are part of a broader reshaping of the checking-account category.
Maybe I’m wrong, but I think a lot of institutions are missing how significant this could become.
For people who do not know MrBeast, because Ron clearly does not, he is one of the biggest YouTubers in the world. His real name is Jimmy Donaldson, and he built a massive audience through high-budget stunts, giveaways, competitions, and related businesses.
But your Array point makes me think about data.
At a technology summit last fall, someone from Baxter Credit Union talked about how they use SavvyMoney.
SavvyMoney provides credit-score and credit-profile insights to members.
The institution may not know the exact lender behind every outside account, but it can see that a member has several credit cards, an unsecured loan, an auto loan, and other obligations.
Baxter uses that information to identify opportunities to serve the member better.
So yes, credit monitoring by itself can be a commodity feature.
The real question is what the institution does with the information.
How prescriptive are you? How do you use data you already have? How do you combine it with other information to create a better offer or experience?
If you deepen the relationship, the account becomes stickier and attrition can decline.
That is the part I think institutions need to focus on.
I love it.
I’m looking at the clock, and we’ve done a good job covering several topics. There are a few more I would have liked to get to, but we’ll save them for next time.
Stacey, thanks as always for your insights, thoughts, and opinions.
And thanks to everybody for joining us on this episode of What’s Going On in Banking. We look forward to seeing you next time.
If you enjoyed today’s episode, follow us on Spotify, YouTube, Apple Podcasts, or wherever you listen.
Thanks for tuning in.
Enjoying What's Going On In Banking?
Subscribe on your favorite platform