Podcast 157: Why your next banker will be an AI, with Brett King

The challenger bank era is largely settled. The interesting question is what comes next.

Brett King has spent fifteen years writing the framework for that question — four books in the Bank series, with the fifth out in November. In this episode of the Fintech Garden Podcast, he walks through what he expects the next ten years to look like.

The summary, in his words: banking becomes like electricity. Embedded in the world. Invisible. Surfaced through whichever AI happens to be managing the user’s moment.

The implications are concrete, and they matter for anyone building, selling into, or running a financial product today.

The Challenger Banks Already Won

The numbers are no longer ambiguous.

The top 20 retail fintechs now have roughly 4.1 billion customers between them. Up around 200% in five years.

The top 20 traditional retail banks grew customer numbers by 3% over the same period.

Nubank has added more than 25 million customers in a few years. Brett expects it to be the largest retail bank in Europe by the end of the decade, possibly sooner. WeBank, out of China, is already the world’s largest digital bank.

This is no longer a story about future disruption. It is a story about what has already shipped.

Why They Won

Three operational advantages separated digital from the incumbents. They are worth being specific about, because they remain the benchmark.

Customer acquisition cost. WeBank reportedly under $1. Nubank in the $5–7 range. Revolut roughly double that. Traditional retail banks operate in the hundreds of dollars.

UX framing. Digital players design around the user’s relationship to their money — savings progress, financial health, balance trajectory. Incumbents still design around products: credit card, mortgage, loan. The cosmetic differences in the app are easy to copy. The framing isn’t.

Onboarding. A Revolut or Nubank account opens in minutes. An SME account at a European retail bank can still take weeks. Internet banking passwords arriving in Germany by physical post in 2026 is not a hypothetical.

These are operational gaps, not branding gaps. That is why they have not closed.

The App Disappears

Here is the central claim.

Within five to seven years, most users will hand their day-to-day banking to a personal AI.

The interaction shifts from app-driven to conversational and generative. “Send Igor €50, what’s the best way to do it.” The AI picks the currency, the rail, and the timing. If a stablecoin route is cheaper and settles faster, it routes through a stablecoin. The user does not see that decision, and does not need to.

Where an interface still exists, it is liquid and contextual. A UI generated in response to a conversation, not a fixed screen the user navigates.

The banking app, as a product surface, is on a timer.

It will not vanish overnight. But the share of meaningful interactions happening through fixed app screens drops sharply over the next five years.

Credit Becomes Contextual. Products Dissolve.

The cleanest example Brett gives is the grocery store.

You walk in. You start putting items in a basket. Your AI notices you don’t have enough in your spending account. It asks: small line of credit, or move money from savings? You answer. The transaction completes.

There is no card. There is no application. There is no product called “consumer credit.”

The underlying utility — short-term liquidity at the moment of need — is delivered. The packaging that bank product teams have built for forty years dissolves into context.

The same logic applies to car finance. Mortgages. Savings products. Cross-border transfers.

Bank products, as discrete things customers shop for, were a distribution mechanism for the branch era. They are not what users actually wanted. They are what users had to accept to get what they wanted.

Once distribution moves into AI, the packaging becomes a tax on the experience.

Brett’s Framework, Updated

Brett’s own categorization is useful as a map.

Bank 1.0 — traditional banking. Bank 2.0 — self-service. Internet banking. Bank 3.0 — mobile. “Banking is no longer somewhere you go, it’s something you do.” Bank 4.0 — embedded. “Banking everywhere, never at a bank.” Bank 5.0 — agentic. The new book, out in November.

Each transition moves the bank further from the customer relationship and closer to the role of underlying utility.

Bank 5.0 completes that arc. The bank does not own the relationship. The AI does. The bank is a utility supplier behind it.

This is the disruption traditional banks have been most reluctant to model. It is also the one that maps most directly to where the technology is actually heading.

AI Is Already Smarter Than Most Humans on Most Measurable Tasks

The conversation does not soft-pedal where AI capability actually sits.

On the GPQA Diamond benchmark — a structured academic reasoning test — leading LLMs now outperform the majority of humans. Only highly specialized PhDs score higher, and only in their own domain.

AI is solving math proofs that humans have not solved.

Where AI is still meaningfully behind is real-world conceptual reasoning — the contextual common sense that does not show up in benchmark sets.

But the disruption AI produces does not depend on resolving philosophical questions about consciousness or hitting a specific AGI threshold.

Jobs are being displaced. Industries are being restructured. That is not a future event.

Recursive Self-Development Is the Inflection

The more important technical signal sits inside the AI development cycle itself.

Roughly half of the new LLM codebase is now written by AI. Within a few years, the expectation is closer to 100%.

Major model releases used to come at roughly 24-month intervals. That cadence has compressed to monthly or bi-monthly.

The combination — AI writing the code that builds the next AI, on a faster release cycle — produces what Brett calls recursive self-development.

Whether or not the result is what anyone agrees to call AGI, the improvement curve becomes effectively exponential.

And the investment behind it now makes any meaningful slowdown structurally implausible.

This is not the part of the future that can be opted out of.

The Payments Rails Are Realigning Globally

Real-time payment rails are the fastest-growing payment infrastructure in the world.

UPI in India. Pix in Brazil. Mobile payment infrastructure in China.

These rails are already more advanced than anything the US has at consumer scale. The US is roughly a decade behind on real-time payments, a consequence of competitive fragmentation among private operators.

On top of real-time, stablecoins have moved into trillions of dollars of annual transaction volume — the verified 2025 figure from Artemis Analytics is around $33 trillion, more than double the prior year. USDC and USDT dominate the supply.

CBDCs are progressing in parallel. China’s e-CNY could plausibly become an interoperability layer for BRICS-nation transactions.

The endgame Brett sketches: AI-mediated payment gateways abstract currency choice from users entirely. Over time, AI agents may develop new mechanisms of value exchange that look unfamiliar by today’s standards.

By the 2040s, the utility of money in its current form may have declined significantly. The more automation runs the economy, the less the current form of money is needed.

This is the part of the conversation most likely to age fastest. Which is exactly why it matters.

Trust Has Moved From Brand to Utility

The traditional case for trusting a bank — physical branches, banking license, brand recognition — no longer maps to how most users actually decide.

Most Revolut customers did not know, until recently, that Revolut lacked a full UK banking license. The fact that Revolut had no branches never appeared in the decision.

Alipay enjoys higher trust scores in China than ICBC or Agricultural Bank of China.

The reason is consistent and operational: utility, delivered reliably, at scale.

In the AI era, trust becomes a shared function across layers — the operating system, the personal AI, the bank, the payment network, the security stack. Whichever layer fails most visibly loses trust first.

For traditional banks, the implication is uncomfortable. The brand they built was an asset that depreciated faster than they noticed.

The New Privacy Is Data Control

Privacy, as a concept, is shifting from secrecy to control.

The question is no longer just who can see your data. It is who can use it, for what, and whether you get paid when they do.

Brett references Drum Wave, a Silicon Valley company building what it calls a data wallet and a data savings account — letting users hold, monetize, and route the data they generate, the same way banks hold, route, and generate yield on money.

This is a useful pattern for fintech. A utility most users will not think about. But one that quietly determines outcomes.

What This Means for Operators

The episode does not deliver a list of action items. It delivers a framework. Pulled out into operational implications, four things stand out for anyone building in fintech now.

The interface layer is not a moat. The app is a temporary product surface. Investment in app UX should continue, but with the recognition that the share of meaningful user interactions through fixed app screens declines over the next five years.

Banking products are not the product. They are a packaging of underlying utility. The product roadmaps that survive are the ones organized around utilities — liquidity at the moment of need, optimal value transfer, and financial visibility — not around legacy product categories.

Trust is a function of utility, not brand. Investment in reliability, transparency, and visible operational performance compounds. Investment in brand alone does not.

Data is the next layer to architect for. The companies that figure out the user side of data — control, monetization, portability — will own a layer that fintech mostly does not own yet.

Brett’s closing point in the conversation is straight: the next ten years of fintech will not be a continuation of the last ten years. It will be a structural shift in interface, product, and trust.

The challenge is not predicting it. It is building for it before it has finished arriving.