When the Cash Machine Arrived: What Really Happened to Bank Staff?

In the average American urban banking market, the number of tellers in a branch fell from 20 to 13 between 1988 and 2004. That looks like a verdict on the profession. Cash machines spread through those years until the United States alone had more than 400,000 of them, according to a 2015 article by the economist James Bessen in the International Monetary Fund's magazine Finance & Development. But Bessen records the other half of the sum: the number of urban branches rose by 43% over the same period.

Why did branches multiply?

Bessen explains that the machine lowered the cost of running a branch, so opening a new one became cheaper, and banks competed for market share by opening more. There were fewer staff in each branch and more branches, so jobs did not fall as far as the first figures suggest. The work itself changed too: less cash counting and more dealing with customers, particularly small-business owners, to whom more profitable services could be sold. One task, handing out cash, was automated, and beside it opened work that the machine does not do.

Then the numbers fell

The story did not end there. The US Bureau of Labor Statistics (BLS) puts the number of tellers at about 339,200 in 2025 and projects a fall to about 294,500 in 2035, a decline of 13%. The Bureau attributes this to the growth of online banking, such as depositing cheques by phone; to shrinking branch networks; to video kiosks that serve more customers with fewer staff; and to enhanced cash machines expected to take over issuing cards and detecting counterfeit notes. The technology that did not reduce jobs in its first phase was followed by another generation with a different effect. The balance was a phase, not a law.

What decides the direction?

This is the editors' reading of the two stories, not a conclusion stated in either. Three questions seem to explain the difference. First, does demand rise when the cost of a service falls? In the first phase the branch network grew because running it became cheaper. Second, is there still something in the work for which people need a human being? What remained was the relationship with the customer, persuasion and responsibility for the customer's situation. Third, which generation of technology is it? What spared the clerk in front of the first machine did not spare them in front of the screen.

And artificial intelligence?

A study by the International Labour Organization (May 2025) places about one worker in four worldwide in an occupation exposed to some degree to generative AI. Some 3.3% of employment falls in the highest exposure categories, led by clerical occupations, which are closest to the processing of transactions and information. The study judges that the dominant effect is likely to be the transformation of jobs rather than their elimination, because most occupations include tasks that need human input. Shares differ between countries: exposure covers 34% of employment in high-income countries against 11% in low-income ones. Exposure measures possibility; it is not a ruling that jobs will be lost.

What should one do?

Ask three questions about your own work. Which of its tasks are routine enough for a tool to take over? What will open up beside them if time is freed, and can I master it? And what does my employer plan for those whose tasks change? The worker should not carry the burden of answering alone: announcing a retraining plan before decisions to cut posts is part of an employer's trust. For the logic of tasks, see the article "A Job Is a Bundle of Tasks: Which Occupations Fade and Which Change?", and for skills, "The Skills That Stay Valuable When Machines Master the Work". Provision rests with God, and striving and learning are among taking the means.