Summary
AEI piece drawing on economist James Bessen’s research to challenge the naive “machines steal jobs” narrative. ATMs reduced tellers per branch from 21 to 13 — but simultaneously made branches cheaper to open, so banks opened more branches, and the total number of tellers increased since 2000. The same dynamic appears in cashiers after barcode scanners, paralegals after e-discovery software, and weavers after 19th-century textile automation.
Key Points
- The ATM paradox: ATMs cut tellers per branch by ~38%. But cheaper branches → more branches. Net result: teller employment grew faster than the labor force from 2000 onward (data through publication date). The BLS projects 8% decline over the next decade — not 80%.
- Bessen’s general pattern: Across many industries, labor-saving technology changes the job rather than eliminating it. Tellers shifted from cash-handling to “customer relationship team” — higher interpersonal skill, more college graduates hired, higher wages.
- Historical parallel: 19th century textile automation mechanized almost all weaving tasks, yet weavers continued to grow for decades as lower cloth prices drove higher demand.
- The mechanism: Technology lowers unit costs → lower prices → higher demand → more total output → more workers needed at the new output level. This is Baumol’s productivity paradox in reverse — automation can create demand-driven job growth.
- The limit: “This doesn’t mean the numbers of bank tellers will continue to grow forever.” BLS then projected 8% decline as online/mobile banking reduces branch visits. The technology eventually catches up — but on a longer timeline than assumed.
Newsletter Angles
- Mechanical Turk Pattern counter-narrative: The ATM story is the optimistic case the Mechanical Turk Pattern can’t explain. Ghost work (hidden humans behind AI) is the pessimistic version; ATMs are the optimistic version. Both are real — the question is which dynamic dominates for any given technology.
- Dynamic pricing and AI parallel: If AI dynamic pricing lowers consumer costs and increases volume (like ATMs lowered branch costs and increased branch count), does it create more service jobs? Or does it increase capacity without adding jobs? The ATM story is a framework for thinking through these second-order effects.
- LLM as teller replacement: Are LLMs to knowledge workers what ATMs were to tellers? The optimistic reading says knowledge workers will shift to higher-value interpersonal and judgment tasks. The pessimistic reading says cognitive tasks are different — language models don’t just handle transactions, they handle the interpersonal parts too.
Entities Mentioned
No specific entities — the article uses ATMs, banks, and tellers as industry-level case study.
Concepts Mentioned
- Mechanical Turk Pattern — the ATM story is the contrast case: visible automation that increases rather than hides labor demand
- Dynamic Pricing AI — the demand-creation mechanism (lower cost → higher demand → more workers) applies to AI pricing optimization analysis
- Leverage Erasure Through Automation — the ATM case is the counter-example: automation raised teller wages and skill requirements rather than eliminating leverage
Quotes
“The impact of the ATM machine was not to destroy tellers, actually it was to increase it.” — James Bessen
“The labor-saving technology actually created more jobs. This is in fact a much more general pattern.”
“What the ATM machine did was effectively change the job of the bank teller into one where they are more of a marketing person.”
Notes
AEI piece from 2016 — written before the current AI wave. The data is accurate for the period it covers. The BLS’s then-projected 8% teller decline has likely materialized since 2016 as mobile banking grew. The piece is valuable as a framework for thinking about automation dynamics, not as current data.