A technology that sharply reduced the number of bank teller workers in the United States was not the ATM but the iPhone and mobile banking, a report said. On July 5, online outlet Gigazine reported that writer David Aks (데이비드 옥스) said the ATM example often cited in AI employment debates has limits in explaining today’s job declines.

Aks said ATMs did not immediately push bank tellers into mass unemployment. He said that was because ATMs replaced only some tasks, such as cash withdrawals, rather than the entire range of banking work.

ATMs, commercialised in the 1970s, spread quickly after Citibank adopted them on a large scale in 1977. The cost per transaction was 27 cents, lower than $1.07 for a teller, and customers could use them outside business hours.

After ATMs spread, the number of tellers per branch in urban areas fell from 21 to about 13. But David Autor said lower branch operating costs and deregulation led the number of urban branches to rise by more than 40 percent from 1988 to 2004. As a result, staffing per branch fell but the total number of tellers increased. The role of tellers also shifted from handling cash to advising on credit cards, loans and investment products.

But that explanation no longer fit the current situation. Full-time bank tellers in the United States fell to 332,000 in 2010 from 235,000 in 2016 and 164,000 in 2022.

Aks cited the spread of smartphones, symbolised by the iPhone unveiled in 2007, and the expansion of mobile banking as the cause of the decline. As bank apps enabled balance checks and cheque deposits and card payments and Apple Pay also spread, customers made fewer trips to branches for tasks they used to hand to tellers.

The number of branches also fell. The number of U.S. commercial bank branches per capita dropped by about 30 percent after peaking in 2009, and Bank of America closed about 40 percent of its branches from 2008 to 2025. The company’s chief executive told CNBC in an interview that the iPhone let customers carry a bank branch in their pocket.

Aks also linked the example to the AI employment debate. He said that if AI is deployed to replace people while keeping existing work processes unchanged, it may only streamline some tasks, like ATMs. He argued that bigger change happens not when AI simply replaces a person’s job, but when work processes are reorganised on the assumption that AI is built in.

The example shows that even with the same automation technology, the impact on employment can differ depending on whether it merely supports existing counters or changes the path customers use to access services. It left the question of whether the AI debate should look beyond replacing individual tasks to whether work structures are reshaped.