For the last ten years or so, banks have spent a lot of time thinking about how to make themselves digital.
The whole idea of the digital bank was to industrialise banking through digital channels. We were essentially doing the same things that banks had done for many years, but digitising them so that they could be done faster, more cheaply and more efficiently.
AI is not a continuation of that.
It is a whole new ball game.
One of the most important differences is that AI does not just help us understand data. It is increasingly able to create meaning from data and then take action from that meaning.
In other words, the data is now able to act.
That changes the nature of the bank itself.
In a traditional institution, we understand the hierarchy. We know where responsibilities sit, who owns a process and who is accountable for an action.
In an AI bank, those responsibilities can start to move.
Agents may make decisions further down the hierarchy. They may interact with other agents and increasingly they may also interact with agents created by customers themselves.
This changes the relationship between the bank and the customer.
In digital banking, the app became the point of engagement. In China in particular, the app and the super app became highly developed platforms through which the institution related to the customer.
In AI banking, I believe the app becomes the most irrelevant part of that interaction.
The customer does not even need to look at your app. He can create his own agent that talks to your institution and goes past the app to relate to you directly.
The power in the relationship begins to turn around.
The institution no longer determines entirely how the customer will interact with it. The customer increasingly decides how he wants to relate to the institution.
Then there is programmable money.
China started with the e-CNY. Around the world, tokenised deposits and tokenised payments are becoming increasingly important. A token can carry value together with information about the transaction.
Now ask what happens when a consumer puts an agent application on a programmable token to make a payment to another end user.
Where does the bank’s relationship come into that transaction?
This is why we cannot look at AI, programmable money, agents and payments in isolation. They are developing at the same time and together they can fundamentally change the way financial transactions take place.
It also creates new questions about governance.
Banks are accustomed to knowing where governance sits. But if agents further down the hierarchy are taking responsibility for transactions, governance has to move down the ranks as well.
And yet the final responsibility still sits with the board and management.
Speed makes this even more important.
We already had a glimpse of what speed can do with Silicon Valley Bank, where a digital bank run unfolded in a matter of days.
Now imagine what a bank could look like when it can become insolvent instantaneously because of AI.
At what point do you use judgement to stop a transaction?
Who holds the kill switch?
These are some of the risks banks will have to work through as AI becomes a more important component of the institution.
Many banks today are proud of the number of AI projects they have under way. An average bank may have 200 to 500 projects at any one time. A lot of these are focused on productivity.
But I think they really need to be focused on the transformation of processes.
The more fundamental question is whether AI changes the product, the institution and the relationship with the customer.
If the product does not change, nothing has changed.
At the Finance Beijing Roundtable, I explored these questions around agentic banking, programmable money, regulation, governance and the changing relationship between the institution and the end user.
The rules are still evolving and both banks and regulators are learning their way through them.
Watch the full speech for my discussion on what the transition to AI banking could mean for the future of financial institutions.
