What The ATM Paradox Means For AI At Work
AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: What The ATM Paradox Means For AI At Work on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get the latest gadgets delivered free with Prime

  • Fast, free delivery on millions of items
  • Prime Video, Amazon Music and more included
  • Member-only deals all year
Start your free Prime trial Free trial for eligible customers · Cancel anytime
As an affiliate, we earn on qualifying purchases.

TL;DR

The ATM took over cash dispensing, but US teller employment rose modestly from about 500,000 in 1980 to about 550,000 in 2010 as banks opened more branches and teller roles shifted toward customer service. The source argues that AI could follow a similar task-and-job pattern, but it does not establish that AI will create enough new work to offset job losses.

A new analysis published November 20 revisits the ATM’s effect on bank tellers, finding that the machine automated a central task without causing US teller employment to fall between 1980 and 2010. The account, by Thorsten Meyer, says ATMs made branches cheaper to operate and helped banks expand, while teller work shifted toward customer interaction — a historical example relevant to, but not proof of, how AI may change jobs.

The first Barclays cash machine opened in Enfield, north London, on June 27, 1967. It dispensed up to £10 in one-pound notes using a special paper cheque and a personal code; modern plastic bank cards did not yet exist. The machine performed a teller’s most visible routine task, dispensing cash, and its name — automated teller machine — made that connection explicit.

In the United States, ATMs grew from about 100,000 in 1995 to about 400,000 in 2010, according to Bank for International Settlements figures cited in economist James Bessen’s analysis. Yet US teller employment rose modestly, from roughly 500,000 in 1980 to about 550,000 in 2010, as David Autor summarized Bessen’s findings. The comparison covers different start dates for machines and employment, and does not mean teller work was unaffected.

Bessen’s explanation is that banks needed fewer tellers at each branch but could afford to open more branches. The source reports a 43% increase in urban US branches from 1988 to 2004, alongside a fall in the typical number of tellers per branch from about 20 to 13. Tellers also moved away from cash handling toward customer advice and sales, including what banks call relationship banking.

At a glance
analysisWhen: Published November 20, 2026; historical…
The developmentA November 20, 2026, article revisits ATM adoption and teller employment to explain how automation can change jobs without immediately eliminating them, and what that history may suggest about AI at work.
Thorsten Meyer AI
Work After Work · Ep. 18

Infographic 01 · The paradox

Four times the machines, more tellers

ATMs took over the teller's core task. For three decades, banks answered with more branches, not fewer tellers.

×4ATMs in the US, about 100,000 (1995) to 400,000 (2010)
≈ +10%US bank tellers, about 500,000 (1980) to 550,000 (2010)
20 → 13tellers per branch in a typical US city, 1988 to 2004
+43%bank branches in US cities, 1988 to 2004

US commercial bank offices (head offices + branches) · year-end · FDIC

  • 198052,879
  • 198858,937
  • 1995ATMs start to spread fast65,578
  • 200477,302
  • 2012peak88,522
  • 202572,658

Teller and ATM figures: Bessen’s analysis of Census and American Community Survey microdata, as summarised by Autor (2015); ATM counts from the Bank for International Settlements. Branch growth in cities: Bessen (2015), urban markets. FDIC series: FDIC-insured commercial banks in the US states and DC; savings institutions are not included.

Sources: D. Autor, Journal of Economic Perspectives 29(3), 2015; J. Bessen, Toil and Technology, IMF Finance & Development (March 2015); FDIC, Historical Bank Data (commercial banks, year-end).
Powered by Thorsten Meyer AI

What ATMs Changed in Bank Work

The ATM example challenges the simple assumption that automating a task automatically eliminates the job attached to it. A job comprises multiple tasks, and reducing the cost of one part of a service can affect how much of that service businesses provide. In the account presented here, cheaper branches and expanded banking access helped offset the need for fewer tellers at each location.

That is relevant to workplace AI because many systems are designed to automate particular activities — drafting, sorting information, answering routine questions — rather than every part of a role. If a business expands its services or employees take on different tasks, employment effects may differ from what a task-level demonstration suggests. But the ATM history is a case study, not a forecast: it cannot show whether demand for AI-enabled services will grow enough to preserve jobs, or whether new work will be accessible to workers whose tasks are displaced.

The teller numbers also depend on the period being measured. The source reports that US teller employment later declined from about 600,500 in 2008 to 339,200 in 2025, according to the Bureau of Labor Statistics. That longer-term fall cautions against treating the early employment rise as proof that automation has no lasting effect.

Products Worth Considering

Apple MacBook Air 15-inch Laptop with M5 Chip, 16GB Memory, 512GB SSD, Midnight

Apple MacBook Air 15-inch Laptop with M5 Chip, 16GB Memory, 512GB SSD, Midnight

Bose QuietComfort True Wireless Earbuds with Active Noise Cancellation, Bluetooth 5.3, USB-C Charging, IPX4, Black

Bose QuietComfort True Wireless Earbuds with Active Noise Cancellation, Bluetooth 5.3, USB-C Charging, IPX4, Black

As an affiliate, we earn on qualifying purchases.

From Cash Machines to Mobile Banking

The ATM’s introduction preceded a long period in which cash access remained tied to bank branches. Barclays’ archive says that before the late 1960s, customers generally had to visit a bank during office hours to get cash. Germany opened a cash machine in Tübingen on May 27, 1968; a Docutel cash dispenser was installed for Chemical Bank in New York in September 1969, according to the Smithsonian and Computer History Museum.

The source describes a later shift driven less by ATMs than by mobile banking. In 2023, 48.3% of US households with a bank account mainly used mobile banking, while 15.1% mainly used a teller. It also reports that German bank branches declined from 45,467 at the end of 2004 to 16,799 at the end of 2025. Those figures suggest that the ATM-era expansion did not continue indefinitely as customers gained other ways to bank.

The article further says cash machines are now retreating in Germany, China and Japan, but supplies no machine counts or time periods for those claims. Its historical argument is narrower: technology can first reshape the economics and tasks of a service, while later technologies and changing customer habits can alter demand for the service itself.

“Actually rose modestly from 500,000 to approximately 550,000 over the 30-year period from 1980 to 2010.”

— David Autor, summarizing James Bessen’s findings in 2015

Where the ATM Comparison Falls Short

The historical figures do not establish what AI will do to total employment. The source offers no comparable data on AI adoption, affected occupations, new job creation, wages or how quickly employers may reorganize work. Outcomes may vary by industry and by whether businesses use productivity gains to expand services, reduce staffing or both.

Some figures also have different scopes and periods. The ATM count is for 1995–2010, while the cited teller employment comparison is for 1980–2010; the 43% branch increase refers to US cities from 1988 to 2004. The article gives no detailed method for its city branch calculation. Separately, the later teller decline is based on Bureau of Labor Statistics figures, but the source does not explain how changes in job classifications or the financial sector affect that comparison.

It is also unclear from the supplied material how much of the teller employment decline after 2008 resulted from mobile banking, branch closures, the financial crisis or other changes. The ATM story supports a lesson about tasks and business demand, but not a settled prediction about AI’s net effect on workers.

The Next Test for AI at Work

The relevant test will be what employers do after AI reduces the time or cost of particular tasks: whether they expand output, redesign roles, cut staffing or combine those responses. Future employment and wage data, alongside evidence on which tasks workers actually perform, will be needed to distinguish a change in job content from a decline in job opportunities.

The source article does not announce a new study or specify a forthcoming research milestone. For now, its central takeaway is historical: the ATM displaced part of the teller’s work, while branch growth and a shift toward customer service helped sustain employment for a time. Whether AI produces a similar pattern remains an open question.

Key Questions

Did ATMs eliminate bank teller jobs?

Not immediately in the United States. The figures cited in the source show teller employment rising modestly from about 500,000 in 1980 to about 550,000 in 2010, even as ATMs spread. Teller employment later fell, reaching about 339,200 in 2025 according to the Bureau of Labor Statistics figures cited by the article.

Why did teller employment rise while ATMs multiplied?

The analysis says ATMs lowered the cost of running branches. Banks used some of those savings to open more branches, even though each branch needed fewer tellers. The work also shifted toward customer advice and sales.

Does the ATM history prove AI will create jobs?

No. It shows one way automation affected a particular industry over particular periods. It does not establish how AI will affect employment, wages or job creation across other industries.

What later change reduced the role of bank tellers?

The source points to mobile banking and the shift away from branch visits. In 2023, 48.3% of US households with a bank account mainly used mobile banking, compared with 15.1% who mainly used a teller.

Source: ThorstenMeyerAI.com

You May Also Like

The rails. Why European agentic commerce is co-defined by two converging regimes.

European agentic commerce is being shaped by two simultaneous regulatory regimes—PSD3/PSR and the AI Act—creating a complex, statutory infrastructure that differs from the US model.

The $60 Billion Bargain: Why Cursor Could Be a Steal for SpaceX

SpaceX announced it will acquire AI coding startup Cursor for $60 billion in stock, a strategic move with potential for high growth and market impact.

The CFO’s new operating system. Anthropic, OpenAI, and the consulting margin that just got compressed.

Anthropic’s $1.5B joint venture and OpenAI’s parallel funding reshape enterprise AI, integrating models into CFO workflows and disrupting consulting margins.

The Cost-Effective Edge Of Claude Opus 5.5 In AI Solutions

Anthropic introduces Claude Opus 5.5, offering top-tier performance at 40% lower costs, with faster processing and improved efficiency for AI applications.