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

By Thorsten Meyer | November 18, 2026 | ThorstenMeyerAI.com

Buying for a business?Offer from Amazon

Get business pricing on tech for your team

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

Work After Work · Episode 16 · Arc 3: Automation in history

It’s Tuesday, eleven in the morning. Your inbox is full, and the week has barely started.

In another version of history, you would be going home right now. Three hours of work a day, fifteen hours a week. That future was sketched in 1930, for your generation, by the most famous economist of his time. So what happened?

Yesterday, in episode 15, we looked at trade: who wins and who loses when countries trade. Today a new part of the season begins: automation in history. We start with the biggest promise automation ever made. More free time.

Key takeaways

  • Keynes got the money right. In 1930 he predicted that living standards in “progressive countries” would be four to eight times higher within a hundred years. By 2022, real income per person was about 4.4 times its 1930 level in Britain and 5.5 times in the United States (Maddison Project, our calculation).
  • He did not get the time. The working week fell fast before 1930 and then stopped at about 40 hours. US factory workers averaged 51 hours a week in 1909, 44 in 1929, and 41.8 in October 2026.
  • The number to remember is 40. Forty hours has been the standard American week since October 1940, when the Fair Labor Standards Act’s overtime line reached its final level.
  • The 15-hour week was a hope, not a dated forecast. Keynes offered “three-hour shifts or a fifteen-hour week” as a way to share the work that would be left.
  • AI saves time, but so far not hours. In August 2026, US workers saved about 2.1% of all work hours with generative AI. A Danish study found no measurable effect of AI chatbots on recorded hours.
Watch the video version of this episode (about 8 minutes). Narration: AI voice clone of Thorsten Meyer.

Products Worth Considering

Kindle Paperwhite Signature Edition 32GB (2024 Model) - Metallic Black

Kindle Paperwhite Signature Edition 32GB (2024 Model) – Metallic Black

As an affiliate, we earn on qualifying purchases.

1930: a bet against the gloom

In 1930, the world was sliding into the Great Depression, and almost everyone was gloomy. John Maynard Keynes was not. “We are suffering just now from a bad attack of economic pessimism,” he wrote, and called it “a wildly mistaken interpretation of what is happening to us”.

His essay had an unusual title: “Economic Possibilities for our Grandchildren”. It was published in 1930 and later collected in his Essays in Persuasion. It is the same essay that gave us the phrase “technological unemployment”, which episode 14 looked at.

But its boldest line was about wealth:

“I would predict that the standard of life in progressive countries one hundred years hence will be between four and eight times as high as it is to-day.”

And from that he drew a startling conclusion: “assuming no important wars and no important increase in population, the economic problem may be solved, or be at least within sight of solution, within a hundred years.” By the economic problem, Keynes meant the struggle for subsistence, which he called “the primary, most pressing problem of the human race”.

The real problem: what to do with the time

If that problem were solved, Keynes thought, humanity would face a new one: “how to use his freedom from pressing economic cares, how to occupy the leisure, which science and compound interest will have won for him, to live wisely and agreeably and well.”

He did not expect people to stop working. “For many ages to come the old Adam will be so strong in us that everybody will need to do some work if he is to be contented.” The old Adam is his phrase for our ancient urge to work and strive. So the work that was left should be shared:

“Three-hour shifts or a fifteen-hour week may put off the problem for a great while. For three hours a day is quite enough to satisfy the old Adam in most of us!”

Read closely, this is not a dated forecast. The four-to-eight-times growth was a prediction with numbers. The fifteen-hour week was a way to spread the remaining work “as widely shared as possible”.

An empty historic factory hall with red brick walls, a wooden floor and old belt-driven lathes under skylights
US factory workers averaged 51 hours a week in 1909 and 44 in 1929. Since 1940, the factory week has stayed close to 40 hours. AI-generated illustration.

The week was shrinking

It wasn’t a crazy idea. Keynes was watching the working week shrink.

The US Census Bureau’s Historical Statistics show the average week of production workers in American manufacturing:

  • 1909: 51.0 hours.
  • 1914: 49.4 hours. That January, Ford cut its working day from nine to eight hours, as episode 5 showed.
  • 1926: 45.0 hours. In September 1926, Ford announced a five-day, 40-hour week for its plants. Its company paper wrote that “Five working days of eight hours, with no overtimes, is to constitute a week’s work”.
  • 1929: 44.2 hours.

From 51 to 44 hours in twenty years. Draw that line forward, and fifteen hours a week a century later does not look so far away.

The law caught up, then the line went flat

In June 1938, President Roosevelt signed the Fair Labor Standards Act, effective 24 October 1938. It set a maximum week of 44 hours in the first year, 42 in the second, and 40 after that, from October 1940. Work above the line was still allowed, but had to be paid at one and a half times the regular rate. In practice, the line defined the standard week.

Germany pushed further, decades later. In 1984, the metalworkers’ union IG Metall struck for seven weeks. The settlement cut the week to 38.5 hours; the union says “the 40-hour wall” was broken. By 1995, working time in the West German metal industry fell step by step to 35 hours.

The number to remember: 40

That’s the number to remember: 40. Forty hours has been the American standard week since October 1940. Eighty-six years later, it is still the line.

The record of factory hours shows how flat it has been:

  • 1940: 38.1 hours (still a weak economy).
  • 1950: 40.5 hours.
  • 1970: 39.8 hours.
  • October 2026: 41.8 hours (production and nonsupervisory employees in manufacturing, preliminary).

The fall that Keynes saw simply stopped. He hoped for fifteen.

Thorsten Meyer AI
Work After Work · Ep. 16

Infographic 01 · The number to remember

The week stopped at 40 hours

Before 1930, the American factory week shrank fast. Since 1940, it has barely moved.

51 hUS factory week, 1909
44 h1929, the year before Keynes's essay
40 hthe US standard week since October 1940
41.8 hUS factory week, October 2026

Average weekly hours · US manufacturing production workers

  • 190951.0 h
  • 192944.2 h
  • 1940Depression aftermath38.1 h
  • 195040.5 h
  • 197039.8 h
  • Oct 2026preliminary41.8 h

How the week got shorter: laws and unions

  • 1926: Ford announces a five-day, 40-hour week for its plants.
  • 1938: Fair Labor Standards Act: maximum week 44 hours, then 42, then 40 from October 1940; above that, overtime pay.
  • 1984-1995: West German metal industry: seven-week strike, 38.5 hours, then 35 hours by 1995.
  • Keynes, 1930: “Three-hour shifts or a fifteen-hour week may put off the problem for a great while.”

1909-1970: series D 803 (BLS data); 2026: BLS production and nonsupervisory employees in manufacturing. Not to time scale.

Sources: US Census Bureau, Historical Statistics of the United States (1975), series D 803; BLS, Employment Situation, Oct 2026, Table B-7 (preliminary); Library of Congress, 40 Hours, 5 Days (Ford 1926, FLSA Sec. 7); US Dept. of Labor, FLSA 1938; IG Metall, Der Kampf um die 35-Stunden-Woche; J. M. Keynes, Economic Possibilities for our Grandchildren (1930).
Powered by Thorsten Meyer AI

The twist: he got the money right

Here’s the twist. Keynes’s growth forecast came true. Using the Maddison Project Database (real GDP per person, in 2011 dollars), our calculation from 1930 to 2022:

  • Britain: about 4.4 times (Keynes’s own country).
  • United States: about 5.5 times.
  • Germany: about 7.4 times.
  • Japan: about 11.5 times; South Korea: about 25 times.

For Britain and the United States, that is right inside his range of four to eight, years before his deadline of 2030. So we got the wealth. We just did not take it as time.

Today: how long we really work

United States. In October 2026, the average private-sector employee worked 34.4 hours a week, part-timers included (BLS, preliminary). Manufacturing workers averaged 40.5 hours.

Germany comes closest to Keynes. In 2025, the average German worker put in about 1,332 hours a year, the lowest figure of all countries in the OECD’s data. The OECD itself warns that its figures suit comparisons over time better than comparisons between countries, because national sources differ. Spread over all 52 weeks, holidays included, that is about 26 hours a week. But this is mostly the result of part-time work and long holidays, not of a short full-time week.

For comparison, hours actually worked per worker in 2025 (OECD):

  • Germany: 1,332
  • Japan: 1,598
  • OECD average: 1,736
  • United States: 1,800
  • South Korea: 1,833

Asia shows the other end. South Korea cut its legal maximum week from 68 to 52 hours with an amendment to its Labor Standards Act in 2018, phased in by company size. In China, employees of enterprises worked 48.3 hours a week on average in September 2026, according to the National Bureau of Statistics. In August 2021, China’s Supreme People’s Court and its labour ministry published a set of model overtime cases and declared the “996” schedule illegal: 9 a.m. to 9 p.m., six days a week, as reported by Fortune and the Global Times.

Two empty wooden deck chairs on a quiet lakeside jetty in soft morning light
Germany comes closest to Keynes, mostly through part-time work and long holidays. AI-generated illustration.

Why the 15-hour week never came

Economists have argued about this for decades. Three explanations fit the evidence best.

1. Our needs grew with our income. Keynes himself left a clue. He split human needs into two classes: needs that are absolute “in the sense that we feel them whatever the situation of our fellow human beings may be”, and relative needs, “those which satisfy the desire for superiority”. The relative ones, he wrote, “may indeed be insatiable; for the higher the general level, the higher still are they.” My reading: much of the extra wealth went into bigger homes, more travel and more services. And each of those is somebody’s job.

2. We did take more free time, just not as shorter weeks. Valerie Ramey and Neville Francis built a measure of time use for the whole American population back to 1900. They found that about 70% of the decline in hours worked was offset by more time spent in school, and that time spent on housework is slightly higher than in the early 20th century. Their surprising result: “leisure per capita is approximately the same now as it was in 1900.” Their measure counts everyone, children and retirees included, so it is not the same as the working week of a full-time worker.

3. The rich did not lead the way into leisure. Keynes expected the wealthy to be the “advance guard” of the age of leisure. The opposite happened. Dora Costa found that in the 1890s, “the most highly paid worked fewer hours than the lowest paid”, but “by 1991 the highest paid worked the longest day”. Peter Kuhn and Fernando Lozano showed that the share of American men regularly working more than 48 hours a week rose between 1979 and 2006, most among highly educated, highly paid and salaried men.

Keynes’s two conditions did not hold either: there was another world war, and the world’s population kept growing.

Myth check: “Keynes predicted we’d all work 15 hours a week by 2030”

Verdict: partly true.

  • The growth part is true, and it came true. He predicted, with numbers, that living standards would be four to eight times higher in a hundred years.
  • The fifteen-hour week had no date. He wrote that three-hour shifts or a fifteen-hour week “may put off the problem for a great while”.
  • It was a way to share work, not a forecast. The idea was to make “what work there is still to be done” as “widely shared as possible”.

A hope, not a prediction.

The promise is back, this time with AI

Now the question is back, because of AI. Will it finally deliver the shorter week?

Time saved: real, but small. The Generative AI Adoption Tracker, run by economists at the St. Louis Fed, Vanderbilt and Harvard with a nationally representative US survey of working-age adults, found in August 2026 that 40.5% of employed respondents used generative AI for work, and that the time saved equalled 2.1% of all work hours. For a 40-hour week, that is about 50 minutes (our calculation). Users alone reported larger savings: in November 2024, 5.4% of their work hours, about 2.2 hours a week.

Hours worked: no change so far. Anders Humlum and Emilie Vestergaard linked surveys on AI chatbot use in Denmark to official labour market records. They estimate “precise null effects on earnings and recorded hours at both the worker and workplace levels, ruling out effects larger than 2% two years after the launch of ChatGPT”. What changes instead is the work itself: new tasks and reorganised jobs.

So far, the time AI saves has not turned into free time. That is the Keynes pattern, so far.

Thorsten Meyer AI
Work After Work · Ep. 16

Infographic 02 · The twist

We got the wealth, not the time

Keynes's growth forecast came true years early. The short week did not.

4-8×Keynes's forecast for living standards in 100 years (1930)
×5.5real income per person, USA, 2022 vs. 1930
1,332 hhours per worker in Germany, 2025: lowest figure in the OECD data
34.4 ha week, US private employees, Oct 2026

Real GDP per person · 2022 vs. 1930 · own calculation

  • BritainKeynes’s country×4.4
  • United States×5.5
  • Germany×7.4
  • Japan×11.5

Hours actually worked per worker · 2025

  • Germany≈ 26 h a week over 52 weeks1,332
  • Japan1,598
  • OECD average1,736
  • United States1,800
  • South Korealegal max week cut 68 → 52 h (2018)1,833

Asia

  • China: employees of enterprises worked 48.3 hours a week on average (September 2026).

Maddison figures in 2011 dollars; historical values are estimates. OECD: hours data suit comparisons over time, not of levels between countries (national sources differ).

Sources: J. M. Keynes, Economic Possibilities for our Grandchildren (1930); Maddison Project Database 2023; OECD, average annual hours actually worked per worker (2025); BLS, Employment Situation, Oct 2026, Table B-2 (preliminary); Korea Ministry of Employment and Labor; NBS China (19 Oct 2026).
Powered by Thorsten Meyer AI

What it means for businesses

  • Decide on purpose what saved time is for. Better work, more output or shorter hours: each is a legitimate choice. If you don’t decide, the time fills up by itself.
  • Measure hours, not only output. If AI makes a team faster but its hours keep rising, the gains are going into more work, not into better work.
  • Time is part of what you offer. In Germany, shorter full-time weeks were won in collective agreements, not handed out by technology. Saved time you give back can be part of the deal.

What it means for you

  • Track the time AI actually saves you. A rough weekly note is enough.
  • Time is part of your pay. If your output stays the same in fewer hours, that is something you can negotiate: fewer hours, more flexibility, or more pay.
  • Expect the line to move only if someone moves it. In the history of working hours, laws, unions and individual agreements shortened the week.

Thorsten’s thesis: what the 15-hour week says about it

My thesis for this season is that automation could become its own sector, up to 90% of the economy within a decade or a few decades. That is my thesis, not a fact, and this season tests it with evidence for and against.

Evidence for it: Keynes’s growth forecast came true, so big leaps over a century are possible. If machines take over most work, his question returns: how to share the work and the income that are left.

Evidence against it, or at least against the idea that more automation means less work for people: for eighty-six years, more wealth did not mean a shorter week. In this history, machines never shortened the week on their own. Laws, unions and choices did. Episode 47 will test the thesis in full.

Thorsten Meyer AI
Work After Work · Ep. 16

Infographic 03 · Why, and what AI changes

Why the short week never came, and why AI hasn't changed that yet

Three explanations from the evidence, and the first data on AI and working hours.

≈ 70%of the fall in US work hours since 1900 offset by more time at school
40.5%of US workers used generative AI for work (Aug 2026)
2.1%of all work hours saved by AI: about 50 minutes in a 40-hour week
≈ 0effect of AI chatbots on recorded hours in Denmark (more than 2% ruled out)

Why the 15-hour week never came · evidence and interpretation

  • Needs grew with income. Keynes: relative needs “may indeed be insatiable; for the higher the general level, the higher still are they.”
  • Free time came in other forms. More years at school; leisure per person about the same as in 1900, by one full-population measure (Ramey & Francis).
  • The best paid work longest. 1890s: the highest paid worked the shortest days; by 1991, the longest (Costa). Long weeks rose most among highly paid men (Kuhn & Lozano).
  • Keynes’s conditions did not hold. Another world war; fast population growth.

AI so far · time saved, hours unchanged

  • USA, August 2026: 40.5% of employed respondents used generative AI for work; time saved = 2.1% of all work hours (self-reported, nationally representative survey).
  • Denmark, 2023-24: AI chatbots: “precise null effects on earnings and recorded hours”, ruling out effects larger than 2% two years after ChatGPT.

Sources: J. M. Keynes, Economic Possibilities for our Grandchildren (1930); Ramey & Francis, NBER WP 12264; Costa, NBER WP 6504; Kuhn & Lozano, NBER WP 11895; Generative AI Adoption Tracker (Aug 2026); Humlum & Vestergaard, NBER WP 33777.
Powered by Thorsten Meyer AI

Limits of this episode

  • Long-run hours are for production workers in US manufacturing (Census series D 803 from BLS data; today BLS production and nonsupervisory employees). Other sectors and countries differ. The 1930s figures are depressed by short-time work in the Depression.
  • International hours come from national sources that differ. The OECD says its data are intended for comparisons of trends over time and are unsuitable for comparing levels between countries in a given year; treat the country ranking as a rough guide. The “26 hours a week” for Germany is our calculation (annual hours ÷ 52).
  • Income per person from the Maddison Project is an estimate in 2011 dollars; historical figures are reconstructions, and Germany’s borders changed. Ratios are our calculation.
  • Surveys: AI time savings are self-reported. The Danish study covers Denmark in 2023-24.
  • China’s 996 ruling is cited as reported by Fortune and the Global Times; we could not open the court’s own page.
  • Interpretations. Why the 15-hour week did not come, the lessons for businesses and for you, and the reading of my thesis are my interpretation.

Series navigation

Sources

  • John Maynard Keynes, “Economic Possibilities for our Grandchildren” (1930), in Essays in Persuasion (W. W. Norton reprint, 1963), pp. 358-373. econ.yale.edu (PDF)
  • US Census Bureau, Historical Statistics of the United States, Colonial Times to 1970 (1975), Part 1, chapter D, series D 802-810 (BLS). census.gov (PDF)
  • US Bureau of Labor Statistics, The Employment Situation — October 2026 (6 November 2026), Tables B-2 and B-7. bls.gov · bls.gov (B-7)
  • Library of Congress, Ellen Terrell, “40 Hours, 5 Days”, Inside Adams (17 January 2024), with Ford News (15 October 1926) and the Fair Labor Standards Act, Sec. 7. blogs.loc.gov
  • US Department of Labor, Jonathan Grossman, “Fair Labor Standards Act of 1938: Maximum Struggle for a Minimum Wage” (Monthly Labor Review, June 1978). dol.gov
  • IG Metall, “Der Kampf um die 35-Stunden-Woche”. igmetall.de
  • Jutta Bolt and Jan Luiten van Zanden, Maddison Project Database 2023. dataverse.nl
  • OECD, average annual hours actually worked per worker (retrieved 9 November 2026). oecd.org
  • Ministry of Employment and Labor (Republic of Korea), Labor Standards. moel.go.kr
  • National Bureau of Statistics of China, “National Economy was Generally Stable with New Achievement Made in High-Quality Development in the First Three Quarters” (19 October 2026). stats.gov.cn
  • “China’s top court deems brutal ‘996’ overtime policy illegal”, Fortune (27 August 2021). fortune.com · Global Times (26 August 2021). globaltimes.cn
  • Valerie A. Ramey and Neville Francis, “A Century of Work and Leisure”, NBER Working Paper 12264 (2006); American Economic Journal: Macroeconomics (2009). nber.org
  • Dora L. Costa, “The Wage and the Length of the Work Day: From the 1890s to 1991”, NBER Working Paper 6504 (1998); Journal of Labor Economics (2000). nber.org
  • Peter Kuhn and Fernando Lozano, “The Expanding Workweek? Understanding Trends in Long Work Hours Among U.S. Men, 1979-2006”, NBER Working Paper 11895; Journal of Labor Economics (2008). nber.org
  • Alexander Bick, Adam Blandin, David Deming and Kerry McKittrick, Generative AI Adoption Tracker (August 2026 wave). genaiadoptiontracker.com · Bick, Blandin and Deming, “The Impact of Generative AI on Work Productivity”, Federal Reserve Bank of St. Louis, On the Economy (27 February 2025). stlouisfed.org
  • Anders Humlum and Emilie Vestergaard, “Still Waters, Rapid Currents: Early Labor Market Transformation under Generative AI”, NBER Working Paper 33777 (2025). nber.org

This article is for information only. It is not career, legal, financial or investment advice.

Researched and drafted with AI assistance, including Anthropic’s Claude.

You May Also Like

Impact of AI Video Tools on Corporate Content Creation and Marketing – (Reference)

Creating compelling corporate content is now easier with AI video tools, but their full impact on marketing strategies remains to be seen.

Are Polymarket Trading Bots Actually Profitable? The Math Behind 2026’s Prediction-Market Arbitrage Industry

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

DeepSWE – The benchmark that made the models spread out again

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

Osaka Data Breach: 1.08 Million Email Addresses Leaked Through a Vendor’s Vendor. What to Do

The City of Osaka confirms that 1,078,713 email addresses of users of its online system leaked after a break-in at Arara, the email delivery service its system provider TKC used. What leaked, how it was found, whether “1 million” is right, what residents should do, and what is still unknown.