There Is No Future for Money as We Think of It Today

For most of my career, the argument about technology and banking has been an argument about the bank. Which channels, which core, which app, which partner. I want to make a more uncomfortable case. The object being made obsolete is not the branch, and it is not the app. It is money itself, as we currently understand it.

I said recently, and I meant it plainly, that there is no future for money as we think of it today. That sounds like a provocation. It is actually just a description of what happens when software starts making financial decisions on our behalf. And that has already begun. We are witnessing the end of human-led decision-making in banking.

Machines Cannot Read Our Money

Here is the problem the industry keeps stepping around. The money we use was designed for humans to hold, count, and move. It was never designed to be read, reasoned about, and acted on by machines at scale. Fiat currency, sitting in an account behind a login and a legacy core, is not natively machine-readable in any meaningful sense. Yet the direction of travel is a world where agents transact, negotiate, and settle continuously, without a person in the loop for each decision.

You cannot run that world on money that needs a human to interpret it. So we will need smart money — and by that I mean the full family: stablecoins, central bank digital currencies, and tokens of various kinds, from deposit tokens to utility tokens. Not as a crypto side-quest, and not as an innovation-lab curiosity, but as the substrate that agentic commerce actually requires. If the medium of exchange cannot be operated by software, the software will route around it, and it will route around you.

Trust Moves From Institutions to Algorithms

Banking has always sold itself on trust. The vault, the brand, the regulator, the human across the desk. I think that definition of trust is quietly being reassigned. Over time, trust in banks converts to trust in algorithms. Not because people stop being cautious, but because they delegate the caution. When your agent consistently gets you a better rate, avoids a fee, and moves your money before you would have thought to, you begin to trust the agent's judgment more than your own — and more than your bank's.

That reassignment is hard for incumbents to accept, because it inverts the thing they are proudest of. And it comes with a corollary that boards find harder still. Control is overrated. Efficiency of capital deployment, the maximisation of returns, and the minimisation of risk are far more critical than the feeling of having a hand on the wheel. The institutions that cling to control as the value proposition will lose to the ones that optimise the outcome.

The Numbers Have Already Turned

If this sounds abstract, look at where customers already are. The top twenty retail fintechs now serve on the order of 4.1 billion customers. The top twenty retail banks serve roughly 2.7 billion. That gap did not exist a decade ago, and it did not open because fintechs had better branches. In the time it took one large incumbent to add 38 million customers, a single challenger added 70 million. The distribution advantage that banking treated as permanent has already changed hands.

This is the same trend I have written about for years — the multi-decade move away from the branch as the primary way people engage with money. What is new is that the endpoint is no longer another app. The endpoint is an agent that holds the relationship, and the money moves to whichever institution's infrastructure serves that agent best. Your customer will not leave you for another bank. They will leave you for an intelligence that happens to settle on another bank's rails.

The Honest Hard Problem Is Ethics

I am an optimist about the capability. I am far more sober about the guardrails. AI is definitely the future of banking, but the real challenge is ethics — and it is hard to code ethical guardrails into AI because we cannot even agree on ethics as humans. We are about to delegate consequential financial decisions, at population scale, to systems whose value judgments we have not resolved among ourselves. That is not a reason to stop. It is a reason to treat the ethical architecture as seriously as the technical one, because whichever values we encode will be executed billions of times without a pause for reflection.

There is a bigger frame around all of this that I do not think banking has fully absorbed. Artificial intelligence is, in my view, the end of capitalism as we know it — not its collapse, but its rewiring, as intelligence rather than labour or capital becomes the scarce input that sets the terms. And the constraint on that future is turning out to be physical: energy, compute, and the geopolitics that decide who controls them. Banking's transformation is a subplot in that story, not the headline. But subplots still end. My honest read is that by the mid-2030s, agentic banking will be as mainstream as neo-banking is today — and the institutions that spent this window defending money as it currently exists will find they defended the wrong thing.

What I Am Asking Executives to Sit With

Stop asking how to add AI to the bank. Start asking whether your money is even legible to the software your customers are already delegating to.

Treat smart money — stablecoins, CBDCs, deposit and utility tokens — as infrastructure you will operate, not a trend you will watch.

Accept that trust is migrating from your brand to the algorithm, and decide whether you intend to be the algorithm people trust or the vault it bypasses.

And put ethics on the build sheet, not the compliance memo. We are encoding judgment at scale. If we cannot agree on the values, the machines will still act on whatever we leave them.