Transcript
Hey everybody. Welcome back to another episode of What’s Going On in Banking, live and in person. Stacey Bryant and I are actually in the same place.
How are you?
It has been great hanging out.
We’re on the island of New Jersey.
New Jersey. What exit?
Somewhere down the Parkway.
We’re in New Jersey for a conference tomorrow, coming straight from another conference in Washington, D.C.
I’ve been working with credit unions since 1999, and this year was my first Governmental Affairs Conference, or GAC.
Nobody could believe it was my first one, but it’s true. Apparently nobody had ever let me go before.
And I know it was your first GAC because you kept calling it GAAC.
I thought it stood for Geeks United for America’s Credit Unions.
Maybe it should.
But yes, we spent the last several days in D.C., and then took the train together to New Jersey.
My overall impression of the conference was that the mood was very positive.
We were at Acquire or Be Acquired a few weeks ago, which was mostly community banks. At GAC, the credit-union crowd also seemed optimistic, but there was a clear sense that the industry knows it has to keep moving.
AI and innovation are no longer abstract future topics. Institutions have to decide what they are trying to accomplish and how they are going to get there.
And a lot of them are realizing they probably cannot build everything themselves, which means fintech partnerships become more important.
That’s what I kept hearing.
I think there was definitely optimism about individual institutions.
People are usually optimistic about their own credit union and what they can control.
But I also sensed caution about the broader economy, regulation, and the future structure of the credit-union industry.
There is an undercurrent of concern about what the industry looks like several years from now.
Speaking of things changing, let’s talk about what’s cracking with Kraken.
There are several quick-hit items before we get to the bigger topic.
First, Kraken’s banking arm received direct access to Federal Reserve payment systems, including Fedwire.
To me, something like this was inevitable.
Other applications from nontraditional players have been rejected over the years, and I’m sure bank lobbying groups will have plenty to say about this one.
But the development reinforces something from the 2026 What’s Going On in Banking report.
For the first time this year, we included crypto providers when we asked banks and credit unions which types of companies they see as competitive threats.
Only about 29% identified crypto providers as a threat.
I thought that number was far too low.
Kraken is another wake-up call.
Robinhood, Chime, Coinbase, Kraken, and other nontraditional players are increasingly reaching into activities that banks once considered safely inside their own domain.
It is changing the plumbing of money.
Payments settle faster, new rails are emerging, and the lines between banking, crypto, and technology keep blurring.
It’s funny that you use the plumbing analogy.
Nvidia’s CEO has talked about how younger people may want to consider careers such as electricians and plumbers because of all the physical infrastructure technology requires.
I didn’t expect “plumbing” to become a banking term too, but here we are.
Second quick hit: stablecoins.
One of the clearest findings in the What’s Going On in Banking survey was how many institutions said they were waiting for more regulatory clarity before doing anything with stablecoins.
Well, the OCC has now proposed rules implementing the GENIUS Act.
The clarity is arriving.
I never loved “we’re waiting for regulatory clarity” as an excuse in the first place, and it holds even less water now.
Institutions need to figure out what they want to do, what they do not want to do, and what their timing should be.
Third, there has been a significant bank-stock selloff tied to questions around credit quality and private credit.
We have a report coming called What’s Going On in Commercial Lending that looks at this issue in more depth.
One finding from that research was that many bankers did not view private credit as a major threat to their own institution.
Their reaction was essentially, “We already do this business. We know our market.”
That may be true operationally, but if you are a public company and investors are pricing private-credit risk into your stock, the threat matters whether or not you personally think it affects your bank.
That’s another wake-up call.
And now the big news item: Block, formerly known as Square, announced plans to cut roughly 40% of its staff, going from about 10,000 employees to roughly 6,000.
The company is framing AI as an enabler of that reduction.
I’m calling BS.
I don’t believe AI is the fundamental reason Block can suddenly eliminate 4,000 jobs.
I think this is a cost-cutting decision designed to improve profitability and appeal to Wall Street.
I have been critical for years of Wall Street’s ability to correctly price technology companies.
Look at what happened with Fiserv. The stock was arguably overvalued for a long time, then the market suddenly corrected after one disappointing period.
Block feels similar in the sense that executives can now point to AI as the story that makes a large headcount cut sound strategic.
A few things bother me about that explanation.
First, you do not cut 40% of an organization in one move without creating enormous internal disruption.
Even if technology can eventually absorb a large portion of the work, removing that many people at once introduces chaos.
Second, I do not believe the technology is mature enough today to replace the range of jobs involved.
There are useful AI agents in the market. Some are generic and some are specific to banking or payments.
But most agents do tasks.
We are starting to see coordination across multiple tasks, but agents are not generally replacing whole jobs unless the job is itself very narrow and repetitive.
And these layoffs are not only software developers.
They include customer support, sales, operations, and other functions.
I am not buying the idea that AI alone suddenly made 4,000 people unnecessary.
Two things come to mind.
First, one of the questions we hear constantly on the road is, “What is the ROI on AI?”
This headline gives executives a very tempting answer: “Look, Block saved money by eliminating thousands of jobs.”
That is a pretty weak way to think about AI return on investment.
Second, this reminds me of an experience I had with Klarna and Affirm.
Last year I bought a Peloton because I travel so much and wanted an easier way to fit workouts into my schedule.
At checkout, I saw both Affirm and Klarna.
I decided to try them for research purposes.
Applying through Affirm was extremely easy. I had one identity-verification issue because of something on my credit profile, and I was able to resolve it with an actual person.
With Klarna, I got stuck in circles with automated support.
I knew exactly what the issue was and what information they needed from me, but I could not get the system to understand it.
Eventually I reached something that was supposed to be human support, and I’m still not convinced I was actually talking to a person.
They couldn’t resolve it either.
I never completed the Klarna transaction. I used Affirm and paid it off shortly afterward.
That experience is exactly why I’m skeptical when companies say agents can simply replace customer-service teams.
Klarna made a similar claim in 2025.
They talked about AI doing the equivalent work of hundreds of customer-service agents.
At the time, I looked at that and thought, “You do not even employ that many customer-service people directly. Much of the work is outsourced. How are you calculating this?”
I reached out and eventually spoke with someone at Klarna.
Their more nuanced explanation was that the goal was not simply to replace a specific number of customer-service agents. It was to change the nature of how they interacted with customers and use agents to help customers make better decisions.
That is a completely different claim.
Fast-forward a year, and the company softened some of the earlier rhetoric around AI replacing customer service.
So here is my theoretical question: what does Jack Dorsey know about AI agents that Klarna did not know?
My answer is nothing.
This is primarily a Wall Street story.
There is one funny footnote.
The original version of my FinTech Snark Tank post on Block got taken down by Forbes about 10 minutes after publication because I said Block had a history of over-hiring.
Someone at Forbes said that was an unsubstantiated claim.
So I rewrote the article with more evidence and caveats.
Then I went online and saw everybody else openly saying, “Of course they over-hired.”
Go back to Twitter.
When Elon Musk took over, he cut a massive percentage of the workforce. The platform did not immediately collapse.
That was not because AI suddenly replaced everyone. It was because the company had hired more people than it ultimately needed.
Block may be dealing with a similar issue.
To be fair, over-hiring can be strategic in Silicon Valley. If you hire the talent, competitors cannot.
But that is not how a community bank is going to operate. No bank is going to hire 500 extra people just to keep them away from the institution across the street.
And there is also a long-term talent issue.
If Block is known for dramatic cuts, what happens when it tries to recruit the next wave of high-quality employees?
People may ask, “Why would I join you if I think I could be eliminated six months from now?”
Exactly.
Now let’s stay on AI agents but bring the discussion back to banking.
In the What’s Going On in Banking survey, we asked specifically about fraud and fraud-dispute resolution.
Roughly 40% of respondents said they were not comfortable using AI agents to support fraud identification, management, and resolution.
I thought that was surprisingly high.
What do you think is driving the discomfort?
Accountability is probably part of it.
If an autonomous system makes a bad decision, who owns the outcome?
The line of business? The model owner? The vendor? The executive team? The board?
It becomes the Spider-Man meme where everybody is pointing at everybody else.
There is also a knowledge gap.
Executives may not fully understand what the agent is doing, what data it is using, where the guardrails sit, and when a human is expected to intervene.
And because fraud is already heavily regulated and high-risk, institutions become even more cautious.
I think that uncertainty creates paralysis, and paralysis gets expensive.
I agree there is a terminology and understanding problem.
Some people hear “AI agent” and think, “If we use this, we are going to have to fire people.”
That is not necessarily how it works.
Agents perform narrow tasks. Increasingly they can coordinate across tasks, but the immediate value is often compressing the time required for a process and bringing the work to a human at a much later or more important decision point.
And I want to distinguish that from the way people casually use “human in the loop.”
Sometimes people say, “The AI will do all the work, but a human will make every decision.”
That defeats much of the point of an agent.
An agent is supposed to make a series of decisions autonomously within defined boundaries.
The banker’s job is to decide which decisions can be delegated and where a human must step in.
That is a process-design problem.
Go back 35 years to business-process reengineering.
The management fad was all about redesigning processes and using technology to eliminate unnecessary steps.
I remember interviewing an executive during an IT-strategy project and asking, “What is your organization’s strategy?”
He said, “We are going to reengineer all our business processes.”
I remember thinking, “That is not a strategy. Put down the Tom Peters book and tell me what the business is actually trying to accomplish.”
But the underlying idea was valid.
Processes were poorly designed and poorly documented, and technology could make them faster and more effective.
We are now in the next phase of that story.
Machine learning, generative AI, and agentic AI give institutions new tools for redesigning processes.
The key word is redesigning.
You cannot simply drop an agent into an existing bad process and assume everything becomes better.
That is why I keep talking about pre-building capabilities.
Before a bank buys dozens of agents, it should define policies around autonomy, data access, decision rights, customer agents, vendor selection, pricing, and return on investment.
Commercial customers are going to use agents too. Their ERP, accounting, and payment platforms will increasingly include agents.
That means banks will have agent-to-agent interactions with customer systems.
What kinds of agents are your customers using? What are those agents authorized to do? What happens when pricing moves from per-seat licensing to per-decision or per-output models?
Do you buy specialized agents from emerging vendors now and integrate them into your loan systems and digital platforms, or wait two or three years for the largest vendors to catch up?
Those are the questions institutions should be wrestling with today.
“We’re not comfortable with agents” is not a strategy.
And answering those questions requires the whole organization.
It is not just the CEO or CIO.
Every business line has different objectives and risks.
Compliance, risk, operations, lending, fraud, technology, and product teams all need a role in the conversation.
We recently had a discussion with a group of credit-union compliance and fraud executives around exactly these issues.
I remember asking you on the train whether you knew all the answers, and you said, “It depends.”
I know the answers. They just depend.
Fair enough.
We have about five minutes left, and there’s one more conversation from GAC I want to bring up.
Every year before the main conference, Dr. Brandi Stankovic and Susan Mitchell host the Underground Collision event.
It is designed to surface the real issues credit unions are dealing with rather than the polished conference version.
One panel focused on the GENIUS Act.
Steve Bohanon from Alkami was moderating, and he asked, “Would you be a genius to act on the GENIUS Act?”
A CEO of a midsize credit union acknowledged that changing money movement represents an opportunity, but then said their members are not asking for stablecoins.
My inner Ron Shevlin wanted to jump out and say, “Let me tell you why that is the wrong question.”
How does a CEO actually know members are not asking for the capability?
They may not use the word stablecoin, but that does not mean they do not want the underlying benefits.
Exactly.
At this point, the average consumer should not care what a stablecoin is.
What they want is faster money movement, faster access, lower cost, more convenience, and better functionality.
Nobody asked Apple for an iPod.
But if you had asked consumers, “Would you like a small device that holds thousands of songs and gives you instant access to them?” a huge number would have said yes.
The need came before the product label.
Saying “our members aren’t asking for stablecoins” often becomes an excuse for not doing anything.
And if the average age of your membership is very high, of course they may not be asking about emerging digital assets.
The more important question is what the 25-year-old you are trying to attract expects.
Younger consumers are already putting money into crypto. A credit union can say, “We don’t think they should do that,” but refusing to offer relevant capabilities may simply make it impossible to acquire them as members.
I have a theory I shared with a credit union recently.
Younger consumers who actively choose a credit union or community bank may actually be more deliberate than people who simply default to Chase or Bank of America.
They may have done more research and chosen that institution for a specific reason.
That creates an opportunity.
Give them better investment tools, credit-management capabilities, faster money movement, and a stronger product experience, and you may earn a much deeper relationship.
Think about all the places consumers now park money: Starbucks, E-ZPass, health-spending accounts, wallets, marketplaces, investment apps, and other digital ecosystems.
To understand how much money you have where, you often have to open five or eight different apps.
What if a bank or credit union actually helped you see all of it in one place and gave useful guidance?
Maybe the institution notices that you keep $100 sitting in E-ZPass even though you rarely use toll roads and could safely keep less there.
E-ZPass itself has no incentive to tell you to take money out.
A financial institution that acts as an advocate for the customer does.
That brings us back to predictive analytics and the use of outside data.
I have mentioned before that some credit unions use SavvyMoney data to understand members’ external credit lines and target them with refinancing or consolidation offers.
The broader question is: where does Stacey have her financial life, and how can the institution use available data to serve her better?
If you do that well, the relationship becomes stickier throughout the customer’s life cycle.
That’s a good place to stop.
Thanks everybody for joining another episode of What’s Going On in Banking. We look forward to seeing you next time.
If you enjoyed today’s episode, follow and subscribe wherever you listen, whether that’s Spotify, Apple Podcasts, or YouTube.
We’ve got more witty, gritty conversations coming your way.
Stay tuned.
Enjoying What's Going On In Banking?
Subscribe on your favorite platform