Banking in the Age of AI: Who Makes the Decision?
For generations, banking was built around a relatively simple relationship.
People trusted institutions with their money. Institutions assessed risk, provided credit, processed payments and offered financial advice.
Technology supported that relationship, but human beings remained clearly visible within it.
Artificial intelligence is beginning to change this balance.
The transformation is not simply about faster banking, better apps or fewer visits to a branch. AI is moving deeper into the systems through which financial institutions understand customers, identify risks and increasingly support decisions.
That makes the present transition fundamentally different from earlier digital change.
From automation to anticipation
Digital banking largely allowed people to perform familiar tasks through new channels.
AI goes further.
It can analyse enormous volumes of information, identify patterns invisible to human observers, detect unusual transactions, process documents and assist employees in making complex assessments.
More importantly, it can anticipate.
A financial institution may increasingly be able to recognise a customer’s emerging needs before the customer expresses them. It may identify unusual spending patterns, possible financial stress or suspicious activity almost immediately.
This can make banking safer and more efficient.
It can also make it extraordinarily personal.
And that raises an important question.
At what point does helpful knowledge become excessive knowledge?
The value — and power — of data
Banks have always possessed sensitive information about their customers.
AI dramatically increases the value of that information because separate pieces of data can now be analysed together, continuously and at enormous scale.
This creates opportunities.
A customer could receive services better suited to individual circumstances rather than being placed into broad categories. Fraud could be identified earlier. Financial difficulties might be recognised before they become crises.
But data does not simply create knowledge.
It creates power.
If an institution knows how we earn, spend, save and borrow, and increasingly has the ability to interpret patterns within that behaviour, questions of privacy become inseparable from questions of fairness.
Customers need more than cybersecurity.
They need confidence that their information is being used responsibly.
They need to understand what information matters in decisions affecting them.
And they need meaningful ways to challenge those decisions.
The algorithm said no
Consider a simple possibility.
A person applies for credit and is rejected.
In traditional banking, that person could reasonably ask why.
What happens if the answer becomes: the system decided?
An algorithm may process thousands of variables and discover relationships that would be impossible for a human analyst to identify. That can improve decision-making.
But complexity can also make decisions harder to explain.
This is particularly important because algorithms are not created outside society. They learn from information produced by societies, institutions and people.
Historical data can contain historical inequalities.
If these are absorbed without sufficient safeguards, an automated system may reproduce them while appearing objective simply because a machine produced the result.
Efficiency should never become a substitute for fairness.
Nor should automation become a way of avoiding responsibility.
Someone must remain accountable.
Keeping humans in banking
AI will inevitably change employment across financial services.
Routine administrative work is particularly suited to automation. Documents can be processed more quickly. Information can be summarised automatically. Standard enquiries can be answered without human intervention.
Some roles will disappear or become smaller. Others will change substantially. New ones will emerge.
But the most interesting question is not simply how many jobs AI will replace.
It is what work should remain human.
Financial decisions often arise at important moments in people’s lives.
Buying a home. Starting a business. Losing employment. Managing debt. Planning retirement. Dealing with the financial consequences of illness, separation or bereavement.
These circumstances contain information that may appear in data.
They also contain fear, uncertainty, aspiration and personal context.
A machine may calculate risk exceptionally well.
That does not necessarily mean it understands vulnerability.
Human judgement therefore becomes more important, not less, precisely where decisions become complicated, exceptional or deeply consequential.
The strongest model may not be artificial intelligence replacing people.
It may be artificial intelligence allowing people to do more genuinely human work.
Trust cannot be automated
This brings banking back to something much older than AI.
Trust.
People accept that financial institutions need information about them. They accept that banks must assess risk. Increasingly, they may also accept that algorithms will assist those assessments.
But acceptance depends on boundaries.
Customers should know when important decisions involve automated systems. They should be able to understand the basic reasons behind decisions that significantly affect them. They should have access to human review where appropriate.
Privacy must also be treated as more than compliance with regulation.
The fact that technology makes something possible does not automatically make its use desirable.
AI therefore requires institutions to ask not only:
Can we do this?
but also:
Should we?
A different kind of bank
The bank of the future may know its customers better than any bank in history.
It may detect fraud within seconds.
It may anticipate financial needs.
It may provide services continuously and almost instantly.
It may operate with fewer physical branches and automate much of the work that once required large administrative teams.
But none of these developments alone will determine whether it is a better bank.
That will depend on what happens to the person on the other side of the algorithm.
Does technology give people greater control over their finances or make important decisions less understandable?
Does personalisation improve service or gradually erode privacy?
Does automation remove unnecessary bureaucracy or remove human judgement precisely where it is most needed?
Does AI support employees or simply replace them?
And, above all, when something goes wrong, is there still a person or institution willing to take responsibility?
These are not technical questions.
They are questions about the kind of financial system we want.
Artificial intelligence will become increasingly capable of analysing our financial lives.
The challenge is to ensure that greater intelligence also produces greater responsibility.
The future of banking will not ultimately be determined by how much machines can know about us.
It will be determined by how wisely institutions choose to use what machines know.
- Evelyn Tuma writes on contemporary social and global challenges, with a particular interest in how technology, economic change and public policy are reshaping institutions, communities and everyday life.