Field notes · The future of banking

Banks as Utility Pipes

When money moves itself, the bank that wins is the one you never see. The question is what it carries while it's invisible.

Topic · embedded finance & what's next synthesized from years of notes Read · ~6 min Stance · invisible is fine until the bad month

What I've come to believe

Banking is dissolving into the back end of other people's products. The winners will not own the customer relationship, they will be the regulated, invisible pipe underneath it, plus the one thing software cannot fake: a license and someone to hold accountable when it breaks. The open question is whether that invisible pipe also carries the path for the person who cannot pay.

The future-of-banking conversation is loud with futurism: central bank digital currencies, settlement in milliseconds, AI agents trading with other AI agents. Strip the noise and there is one structural move underneath all of it. The bank slides from the front of the experience to the back, becomes a utility, and has to decide what it is actually for once nobody can see it. That is a more interesting question than any of the gadgets.

01The through-line

For most of history a bank was a vertical column: it owned the experience, the product, the ledger, and the license, top to bottom, inside its own four walls. Embedded finance breaks the column on its side. The experience now belongs to someone else's app. A layer of best-of-breed suppliers handles identity, payments, and compliance. And the bank settles to the bottom of the stack as the regulated layer that actually holds the money and carries the legal weight.

That position is not a defeat. It is a different and durable kind of power, because the regulated layer is the part nobody else can casually replace. But it is invisible by design, and invisibility changes what gets built and what gets forgotten.

The experience moves up, into other people's products. The bank moves down, into the pipe. The customer stops seeing it, which is exactly when the parts no one watches start to matter most.

The column tips on its side. Experience floats up into other people's apps, the bank settles to the bottom as the regulated pipe nobody sees.

02Three reads

Embedded finance demotes the brand. When the experience lives inside someone else's app, three centuries of heritage and a stadium sponsorship stop counting for anything. What counts is the quality of the integration and the price. The logo that took a hundred years to build is worth very little at the back of a stack the customer never looks at.
Atomic settlement removes the float, and the excuses. Real-time, always-on settlement collapses the "where is the money" problem that has defined banking operations forever. It also removes the buffers and delays that institutions have quietly hidden friction and risk inside. When money is final in an instant, there is nowhere left to tuck the slow parts.
The bot has no loyalty. When an AI agent shops for yield on someone's behalf, switching providers is instant and emotionless. It does not care that a bank is old or respectable. The only moats left are being genuinely better on price and performance, or being the one regulated layer that legally cannot be swapped out.

03How I read it

My take

Everyone narrates this future as front-end magic: invisible payments, a car that pays its own tolls, money that just happens in the background. I read it from the back, as usual, and from there the invisibility is the worrying part. When the bank becomes a utility pipe, the human moments do not disappear, they just get harder to find. The person who cannot pay this month is no longer standing in front of a bank. They are stuck inside a merchant's app, three taps from a checkout and zero taps from anyone who can offer them forbearance.

So my contrarian read is simple: embedded finance is wonderful right up until the bad month, and almost nobody is designing the embedded equivalent of "I need to talk to someone because I cannot pay." Frictionless access is easy to build and demos beautifully. The hardship path is invisible work on top of an invisible layer, which means it is the thing most likely to be left out entirely. The whole point of pushing the bank to the back is that it carries the regulated load. Forbearance is part of that load, and it is the part that does not get a launch.

The pipe is the means. The person on a bad month, now buried two layers deeper than before, is still the point.

04Where this is going

Banks survive this. They become the regulated trust layer that feeds the whole ecosystem, the part with the license, the balance sheet, and the accountability that no platform can fake. That is a real and durable role. The open question is not whether the pipe gets built, it will. It is whether the pipe also carries the path for the person who falls behind, or whether frictionless access quietly leaves them stranded inside an app with nowhere to turn. That is the design decision I would be watching, and it is the one nobody is putting on a roadmap.

Threads worth pulling: the "Decoding: Banks" series (11:FS), the episodes on the future ecosystem, embedded finance, and atomic settlement; plus reading on banking-as-a-service and agentic commerce. Claims here are paraphrased from those notes and are my own synthesis.