The Story Nobody Is Telling
This is the story of the ATM and the bank teller — and why it matters for AI. The lesson isn't just that jobs survived. It's that the technology didn't replace the whole role; it took over the repetitive part so that people could take on more and be more productive.
In 1970, roughly 300,000 Americans worked as bank tellers. Then the Automated Teller Machine showed up — a metal box that could do the teller's most common task (dispensing cash) 24 hours a day, seven days a week, without a lunch break, a sick day, or a pension.
By 2010, the United States had deployed over 400,000 ATMs. And the number of bank tellers? It hadn't shrunk. It had doubled to roughly 600,000. The machine didn't replace the human; it changed what the human did — and the human did more of it.
Productivity, Not Replacement
Economist James Bessen explained the paradox in his research for the IMF: ATMs reduced the number of tellers needed per branch from 20 to 13 between 1988 and 2004. But by taking cash-dispensing off the teller's plate, ATMs made each remaining teller more productive — free to focus on opening accounts, advising on products, and handling complex transactions that machines couldn't. Banks could run more branches with fewer tellers per location, and they did: 43% more urban branches. The result wasn't fewer tellers overall; it was more tellers doing higher-value work.
The advancement didn't replace the whole role. It automated the routine slice so the role could expand. Tellers stopped spending all day counting cash and started spending it on relationship banking. Same job title, more output per person, more people in the job.
The Bank Teller vs. ATM Scoreboard
| Year | Bank Tellers (U.S.) | ATMs (U.S.) |
|---|---|---|
| 1970 | ~300,000 | — |
| 1980 | ~400,000 | growing |
| 1990 | ~500,000 | ~100,000+ |
| 2000 | ~550,000 | ~250,000+ |
| 2007 (peak) | ~600,000 | ~400,000+ |
| 2024 | ~347,000 | ~400,000+ |
The table above summarizes the trend; BLS and historical sources (e.g. AEI, Bessen/IMF) provide the underlying employment and ATM counts.
The Second Act
None of this means you should be complacent. The bank teller story has a second act that's less comforting.
The decline that finally came in the 2010s wasn't caused by ATMs at all. It was driven by mobile banking and online account management — an entirely different technology. Teller jobs have fallen roughly 42% from their 2007 peak to about 347,000 in 2024, with BLS projecting another 13% decline through 2034. The threat that eventually arrived wasn't the machine that replaced the task; it was the app that replaced the branch.
The lesson: the technology that looked like the job-killer often wasn't. It made the role more productive first. The disruption that eventually shrinks the job is often something else, years later.
Could AI Be Like the ATM?
That's the theory worth taking seriously. Not "AI will create more jobs than it destroys" — maybe it will, maybe it won't. The more precise idea is: AI might not replace whole roles so much as take over the routine parts so that people in those roles can take on more and be more productive.
- ATMs were supposed to kill bank tellers. Instead they handled cash; tellers handled people. Employment doubled.
- Spreadsheets were supposed to wipe out accountants. Instead they automated the grunt work; accountants did more analysis and advisory. The profession grew.
- The pattern: the advancement doesn't replace the whole. It automates the repetitive slice so the human can do more of the rest.
If AI follows a similar path, then the headline "robots are coming for your job" is the wrong frame. The right question is: what part of your job could the tool do so that you can do more of the part that actually needs you? Productivity gains first; displacement (if it comes) often from a different technology or a different cause.
The Bank Teller Playbook for 2026
Assume the ATM pattern before you assume the apocalypse:
- Expect the role to expand, not just change. The repetitive tasks will get automated. The work that needs judgment, relationship, or context is where you take on more. Tellers didn't disappear; they stopped counting cash and did more advising and selling.
- Use the tool to do more. The goal isn't to compete with the machine on the machine's task. It's to let the machine handle that so you can handle more of what only you can do.
- Productivity first. The ATM made each teller more productive per hour. AI might do the same for your role — same title, more output, possibly more people in the role before any later disruption.
- Watch for the real disruptor. Tellers peaked when ATMs were everywhere. The decline came from smartphones and online banking. The thing everyone panics about often isn't the thing that eventually moves the numbers.
Remember the Teller
In 1970, a machine arrived that could do a bank teller's core job. Over the next four decades, the machine took over the repetitive part — and the human workforce grew. Tellers didn't get replaced; they got more productive. They took on more relationship work, more advisory work, more of the work that couldn't be automated. The role didn't shrink; it expanded.
So the next time a headline says the robots are coming for your career, ask: could this be more like the ATM? Not replacing the whole, but taking the routine so you can take on more and be more productive. The teller is still at the branch. They're just doing more — and different — work than they were in 1970.
Sources: American Enterprise Institute, "What the Story of ATMs and Bank Tellers Reveals About the Rise of the Robots and Jobs" — historical teller employment growth (300K in 1970 to 600K by 2010); James Bessen / IMF research on ATMs and tellers-per-branch; BLS Occupational Outlook Handbook (teller employment and projections).