For the last ten years or so, banks have been trying to become digital.
We built apps, automated processes, cleaned up our data and made banking faster and cheaper.
But AI is not simply the next step in that journey.
It changes the rules.
In my first book, The Great Transition: The Personalization of Finance Is Here, I wrote about how power in finance was beginning to move away from institutions and towards the individual.
My more recent book, Building the AI Bank: Speed, Judgement and Power, takes that argument further. AI does not just change the technology inside the bank. It starts to change how decisions are made, where judgement sits and who ultimately has control.
The big difference is that AI does not just process information. It can increasingly interpret what is happening, make a decision and then act on it.
And once that happens, the relationship between the bank and the customer starts to change.
In digital banking, the app was everything. That was where the bank met the customer.
In AI banking, the customer may not even need to open your app. He may have his own agent talking directly to the bank, deciding what he needs and carrying out the transaction.
That turns the relationship around.
Then we add programmable money and tokenised transactions into the picture. If an agent can carry instructions, make a decision and complete a payment, where does the bank come into that process?
And when something goes wrong, who is responsible?
This is where governance becomes very interesting.
Banks are used to hierarchy. We know who approves what. We know where responsibility sits.
But when agents start making decisions further down the organisation, governance has to move with them.
At the same time, the final responsibility still sits with management and the board.
And then there is speed.
If AI can complete transactions instantly, the transaction may move faster than the governance process designed to control it.
So who holds the kill switch?
Banks today are running hundreds of AI projects, many of them focused on productivity. But I think the more important question is whether AI actually changes the product, the institution and the relationship with the customer.
Because if the product does not change, nothing has changed.
These were some of the questions I raised at the Finance Beijing Roundtable as banks and regulators try to understand what an AI-native financial system might actually look like.
