AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: What Typesetters’ Disappearance Can Tell Us About AI on ThorstenMeyerAI.com

Buying for a business?Offer from Amazon

Get business pricing on monitors, keyboards and dev gear

  • Business-only prices and quantity discounts
  • Tax-exempt purchasing
  • Multiple users, one account, clear invoices
As an affiliate, we earn on qualifying purchases.

TL;DR

A November 21, 2026, essay examines the decline of elevator operators, typists and typesetters to assess what past automation may suggest about AI and work. It draws on economist James Bessen’s finding that only one of 271 occupations from the 1950 U.S. census largely disappeared because of automation, while warning that roles concentrated on a single task can fall quickly.

Elevator operators, typists and typesetters offer contrasting examples of how automation can alter work, according to an analysis published November 21 by Thorsten Meyer. The article argues that machines have generally taken over tasks rather than entire occupations, but jobs centered on a narrow set of tasks can contract sharply—a distinction relevant as employers adopt AI tools.

The analysis draws on economist James Bessen’s study of 271 detailed occupations counted in the 1950 U.S. census and tracked through 2010. Bessen found that only elevator operators largely disappeared because of automation. Other occupations also vanished or changed, but for reasons including shifting demand or the obsolescence of older technology. The distinction matters: an occupation disappearing is not, by itself, proof that automation caused it.

The other examples show how the effects can differ. The U.S. had 985,000 word processors and typists in 1988; the source says the occupation fell by 51% by 2000, far more than the 6% decline official forecasters had expected. It reports about 40,400 such workers in 2025 and says the occupation ranks among the fastest-declining jobs in the country. In Germany, a three-week strike in 1978 slowed the introduction of computer-based typesetting, while the typesetter apprenticeship was merged into a media-design trade in 1998.

The article also points to a broader labor-market measure: office and administrative support accounted for about one in nine U.S. jobs, and the Bureau of Labor Statistics’ August 2026 projection identified the group as expected to shrink fastest by 2035, partly because of AI tools. That is a projection, not a tally of jobs already lost to AI.

At a glance
analysisWhen: Published November 21, 2026; includes a…
The developmentThorsten Meyer’s November 21, 2026, article uses three historical occupations to argue that automation usually reshapes jobs rather than eliminating them, while some task-focused roles can sharply decline.
Thorsten Meyer AI
Work After Work · Ep. 19

Infographic 01 · The number to remember

The one job automation wiped out

Elevator operators rose with the skyscrapers and fell with the automatic elevator. Of 271 occupations in the 1950 census, it is the only one that automation largely wiped out.

97,000US elevator operators in 1950, the peak (census)
37,000in 1970 (census)
1 of 271occupations of the 1950 census that automation largely wiped out by 2010: this one (Bessen)
24,700US elevator installer and repairer jobs in 2025 (BLS), median pay $109,910

US elevator operators · census · thousands

  • 190013k
  • 191025k
  • 192041k
  • 193068k
  • 194087k
  • 1950the peak97k
  • 196077k
  • 197037k

Story vs. evidence

  • Documented (TIME, 8 Oct 1945): 16,600 elevator operators and building workers on strike in 2,100 Manhattan buildings for five and a half days.
  • Documented (Otis, company account): 1948 a recorded voice to help passengers trust automatic elevators; 1950 the first building with Otis high-speed, no-operator elevators (Dallas).
  • Not documented: that the strike made owners switch to automatic elevators. A popular story; no source we found shows it.

Census figures: economically active persons 14 and over, each census in its own classification (1950 in the 1950 classification). Today the BLS publishes no separate count of elevator operators; installers and repairers are a different occupation.

Sources: US Census Bureau, Historical Statistics of the United States (1975), series D 233-682 (row D 587); J. Bessen, How Computer Automation Affects Occupations (2016); TIME, 8 Oct 1945; Otis company history; BLS Employment Projections 2025-35 (27 Aug 2026).
Powered by Thorsten Meyer AI

AI’s Effects on Office Work

The historical comparison cautions against treating automation as a simple choice between workers keeping their jobs unchanged and machines eliminating them. In many occupations, technology has taken over particular duties while leaving other work in place. For AI, that suggests job titles alone may be a poor measure of disruption: changes to the tasks inside a role may arrive before a role disappears.

But the examples also show why some workers may face sharper effects. When most of a job consists of a task that new technology can perform, there may be fewer remaining duties to reorganize around. The typist’s steep decline is a warning about that concentration, though it does not establish that AI will produce the same outcome across today’s office jobs. The BLS projection signals expected pressure on a large occupational group, not a precise forecast for individual workers or a finding that AI alone will cause the decline.

Products Worth Considering

LG UltraGear 34-Inch Curved QHD Gaming Monitor, 160Hz, FreeSync Premium

LG UltraGear 34-Inch Curved QHD Gaming Monitor, 160Hz, FreeSync Premium

Jackery Solar Generator 1000 v2 with 200W Solar Panel, 1024Wh LiFePO4 Battery

Jackery Solar Generator 1000 v2 with 200W Solar Panel, 1024Wh LiFePO4 Battery

As an affiliate, we earn on qualifying purchases.

From Manual Lifts to Digital Text

Elevator operators once controlled lifts by hand, managing doors and stopping at floors. The U.S. census counted about 97,000 operators in 1950, after the occupation grew alongside high-rise buildings. Automatic elevators were available, but passenger confidence was an issue. Otis’s company history says it offered recorded voices for its automatic elevators in 1948 while attendants were still present. In 1950, the company installed an attendant-free Autotronic system in a Dallas office building.

A frequently repeated account links a 1945 New York elevator strike to building owners’ decision to automate. The source says TIME reported that 16,600 elevator operators and building service workers struck in 2,100 buildings over a wage ruling. The walkout lasted five and a half days and ended after Gov. Thomas Dewey called for an impartial arbitrator. Meyer says he found no document showing the strike caused the move to automatic elevators; the connection should be treated as a popular explanation, not established fact.

The subsequent census counts show the occupation declining, from about 77,000 in 1960 to 37,000 in 1970. Bessen’s analysis, as presented in the article, puts that history in perspective: among the 271 occupations he tracked, elevator operator was the sole case largely attributed to automation. The comparison does not mean automation has had little effect on work; it means whole occupations disappearing is different from tasks being automated within jobs.

What the Historical Record Cannot Prove

The link between the 1945 strike and elevator automation remains unverified in the source article: it reports finding no document establishing that the walkout prompted building owners to replace operators. The strike itself is documented, but its alleged role in accelerating automation is not established. The source also notes that later retellings use higher strike figures than TIME’s contemporary report.

The history does not determine how many jobs AI will eliminate, how quickly changes will happen or which workers will move into other roles. The BLS figure cited is a projection through 2035 for office and administrative support, with AI identified as one contributing factor. The source provides no separate estimate of AI’s specific contribution, and the projection does not show how employers or workers will respond.

Watching AI’s Workplace Effects

The next evidence will come from how employers introduce AI and whether it changes specific duties, staffing levels or the mix of roles within offices. The BLS projection through 2035 offers a benchmark for the office and administrative support group, but subsequent employment data will be needed to compare forecasts with actual changes. The source article does not identify a scheduled follow-up report or a new policy decision.

For readers tracking the shift, the historical cases suggest watching both occupation counts and the tasks workers perform. A shrinking job title can signal substantial disruption, but it does not on its own reveal whether work moved to another occupation, changed form or disappeared. The scale and distribution of AI’s effects remain open questions.

Key Questions

Does the article say AI will eliminate office jobs?

No. It discusses a BLS projection that office and administrative support will shrink fastest among occupational groups by 2035, partly because of AI. That projection is not a claim that AI will eliminate the group or that the forecast has already happened.

What does Bessen’s figure of one out of 271 mean?

In the study described by the article, one occupation—elevator operator—was found to have largely disappeared because of automation among the 271 detailed occupations from the 1950 census tracked through 2010. It does not mean automation affected only one occupation; the study’s distinction is between automating tasks and eliminating an occupation.

Did the 1945 elevator strike cause buildings to automate?

The source article says the strike is documented but that it found no document proving the strike caused automation. That causal link remains a popular account rather than a confirmed fact in the material provided.

Why are typists part of the comparison?

The article says the U.S. had 985,000 word processors and typists in 1988, followed by a reported 51% decline by 2000. It uses the example to show how a job built around tasks affected by new technology can contract more quickly than occupations whose duties are broader.

Source: ThorstenMeyerAI.com

You May Also Like

Multi-Vector Embedding Training: A Token Pooling-Aware Approach

A Hugging Face post reports 97.8% NanoBEIR nDCG@10 retention after fourfold vector pooling, versus 94.6% for a normally trained model.

Exploring Muse Spark Open, America’s Strongest Open-Weight Model

Meta released an 816-billion-parameter open-weight model that leads US peers on Artificial Analysis tests, with mixed results on cost and agent tasks.

Teardown: A Generic 7-Port USB 3.0 Hub That Wasn’t

A recent teardown uncovers that a widely marketed 7-port USB 3.0 hub is not what it claims to be, raising concerns over product authenticity and safety.

Facebook Instagram Outage

Facebook and Instagram suffered a widespread outage on April 27, 2024, affecting millions of users globally. Services are partially restored, but some issues remain.