One of the quietest sources of profit for banks is located in the most familiar place to all of us: The money simply sitting in checking accounts.
Customers leave trillions of dollars in deposits and savings accounts, even when higher interest rates are available elsewhere. Moving funds requires comparison, opening an account, and often a fair amount of effort, so most money remains where it is. Artificial intelligence agents threaten to eliminate this exact friction.
Over the weekend, Reuters dedicated its "Chart of the Week" to this new phenomenon: In the US, approximately $3.8 trillion is held in savings accounts. According to calculations, if competition driven by AI agents forces banks to raise average interest rates from about 1.9% to roughly 4%, the additional cost to the industry could reach about $79 billion per year.
Instead of you searching for interest rates, the computer will do it
The scenario is quite simple. An AI agent receives permission to view your accounts, understands how much money is needed for ongoing payments, and identifies that tens of thousands of shekels or dollars are simply standing idle in the account. Simultaneously, it scans banks, money market funds, and financial platforms to find where the money can yield higher returns.
Currently, some agents are still primarily in the recommendation stage. Meta Muse, for example, can connect via Plaid to information at more than 12,000 financial institutions, but it does not yet transfer the money itself. However, the direction is clear. According to PYMNTS, the three largest banks JPMorgan, Bank of America, and Wells Fargo together hold about $1.6 trillion in deposits that pay no interest at all.
If machines begin transferring money automatically every time they find a better offer, the advantage of the "customer who doesn't bother to switch" could disappear.
This week, we already saw an agent that pays on its own.
This is no longer just a theoretical scenario. HSBC and Ant Digital announced in recent days that they completed an experiment in which an AI agent located a digital service, used it, and executed a payment itself using tokenized bank deposits. The clearance took place in real time on a trial blockchain network, with HSBC operating the risk assessment mechanisms.
The experiment is not yet open to customers, but it demonstrates a major leap: Moving from AI that tells us what is worth doing to AI that executes financial transactions on behalf of the user.
Banks can also profit
The irony is that the main threat to banks is also one of their greatest opportunities for efficiency. Even today, banks use AI models in customer service, writing code, fraud detection, anti-money-laundering checks, and credit analysis. Reuters notes that Bank of America, Société Générale, and other banks are already attempting to translate the technology into hundreds of millions of dollars in annual savings.
Therefore, the real battle may be between two types of agents: The bank's agent, which tries to retain the customer, and the customer's agent, which seeks the best deal for them without any brand loyalty.
And what about Israel?
In Israel, a public experiment similar to HSBC's has not yet been published, but the Bank of Israel has already defined agentic AI as one of the major trends expected to reach the banking system, alongside applications in credit, fraud, and risk management. The Banking Supervision Department also found that local banks are still in relatively early stages of implementation.
If agents indeed receive permission in the future to compare deposits, interest rates, and fees and even transfer money with a single confirmation click, the impact in Israel could be particularly significant in a market where many customers still hold large balances in checking accounts.
The revolution, then, is not just another chatbot in an app. It could harm one of the most profitable advantages of traditional banking.