By Thorsten Meyer | November 2, 2026 | ThorstenMeyerAI.com
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
Work After Work · Episode 1 · Arc 1: How the economy is built
Around the year 1800, about three in every four American workers farmed. Today it is about one and a half in a hundred. The food did not run out, and the people did not disappear. They moved: first into factories, then into shops, offices, hospitals and schools. Economists have a simple map for that journey, and it has only three boxes: field, factory and office.
This is the first episode of Work After Work, a daily series of 60 episodes on how the economy is built, how work has moved, and which work stays human as machines take on more of it. We start with the oldest map, because every later question in this season, from the tractor to the chatbot, is a question about people moving between these boxes. And at the end, I put my own thesis on the table, clearly marked as a thesis.
Key takeaways
- Three boxes. The primary sector takes things from nature (farming, fishing, forestry and, in many statistics, mining). The secondary sector makes things (manufacturing, construction, energy). The tertiary sector serves people (trade, transport, offices, health, education, government).
- The map is less than a century old. The New Zealand economist Allan G. B. Fisher added the word “tertiary” in a book written in 1934 and published in 1935. Colin Clark measured the shift across countries in 1940. Jean Fourastié argued in 1949 that services would absorb the workers that machines freed up.
- America left the farm. In Stanley Lebergott’s estimates, farming employed 73.7% of the US labor force in 1800, 54.8% in 1850, 40.2% in 1900 and 12.0% in 1950. The ILO’s modelled estimate for 2025 is 1.5%.
- Asia did it faster. In China, the primary sector’s share of employment fell from 83.5% in 1952 to 22.2% in 2024. In Japan’s census it fell from 53.8% in 1920 to 3.2% in 2020.
- The number to remember: 2001. From that year on, according to the ILO’s modelled estimates, services have employed more people worldwide than farming.
- Myth check: “America doesn’t make things anymore.” False. US factory jobs are about 35% below their June 1979 peak, but real manufacturing output in 2025 was about 1.9 times its 1979 level.
- My thesis, not a fact: within a decade or a few decades, automation could become a sector of its own, doing up to 90% of the work with no human hand. Episode 48 puts it on trial.
Products Worth Considering
Bose SoundLink Plus Portable Bluetooth Speaker, Black
Jackery Explorer 300 Portable Power Station, 292Wh
Amazon Kindle Scribe Colorsoft, 64GB, 11-Inch Color Display with Premium Pen
As an affiliate, we earn on qualifying purchases.
Three boxes: field, factory, office
Take a loaf of bread. Someone grows the wheat: that is the primary sector, the work of taking things straight from nature. Farming, fishing and forestry are always in it. Mining usually is too, but not everywhere: Colin Clark put mining in the secondary sector in 1940, and so does Japan’s census today.
Someone mills the flour and bakes the bread in a plant: that is the secondary sector, the work of making things. Factories, construction and energy supply belong here.
Someone drives the bread to the shop, sells it, keeps the accounts, runs the advertising, and teaches the baker’s children: that is the tertiary sector, services. It is the biggest and most mixed box. Fisher himself called it “a vast miscellaneous residue of activity”, ranging “from transport and commerce through amusements and education to the highest forms of creative art and philosophy”.
The three boxes are blunt. A software engineer at a car maker counts as manufacturing in some statistics and as services in others. But blunt is useful: the boxes show where people earn their living, and how that changes over time.
Who drew the map
The words “primary industry” and “secondary industry” were everyday statistical terms in Australia and New Zealand in the early 20th century, for farms and mines on one side and factories on the other. That is how Colin Clark later described it, in the third edition of his book (1957), where he credits the New Zealand economist Allan G. B. Fisher with coining “tertiary”. Fisher wrote that he had written The Clash of Progress and Security in 1934; it was published in London in 1935. In a 1939 article and again in 1945 he set out the three-way split and the claim behind it: in every progressive economy, he wrote, employment shifts from primary activities “to secondary activities of all kinds, and to a still greater extent into tertiary production”.
Colin Clark, then working in Queensland, Australia, put numbers on the idea in The Conditions of Economic Progress (London, 1940). He compared country after country and called the movement of workers “from agriculture to manufacture, and from manufacture to commerce and services” the most important concomitant of economic progress. He also noticed something that matters for this whole season: the share of industry does not rise forever. It climbs to a maximum and then begins to fall.
The French economist Jean Fourastié gave the model its hopeful twist in Le Grand Espoir du XXe siècle (The Great Hope of the Twentieth Century, Paris, 1949). He sorted the sectors by how fast technical progress moves in them: fast in industry, medium in farming, slow in services. A machine can double a factory’s output; it is much harder to double the output of a hairdresser or a teacher. So, he argued, services would keep absorbing the people that machines no longer needed. A review of the book in the journal Population in 1949 summed up one of his conclusions this way (our translation): the fear of lasting technological unemployment, resulting from the replacement of man by the machine, is in vain.
Hold on to that sentence. This whole season is, in a way, a test of it.
Time travel: how America left the farm
The longest consistent US series comes from the economic historian Stanley Lebergott, published by the US Census Bureau in Historical Statistics of the United States (1975). It counts the labor force aged 10 and over, by industry. Farming’s share of it:
- 1800: 73.7% (1.4 million of 1.9 million; an estimate)
- 1850: 54.8%
- 1900: 40.2%
- 1950: 12.0%
- 1960: 8.1%
Today, the International Labour Organization’s modelled estimate for the United States is 1.5% in agriculture, 18.9% in industry and 79.6% in services (2025). The Bureau of Labor Statistics puts agriculture at 0.9% of total US employment in 2025, but that figure covers wage and salary jobs only, and many farmers are self-employed.
Two cautions. The 1800 figure is an estimate built from scattered records; in the same table, 1810 is even higher (83.7%), which tells you how rough the early numbers are. And the early labor force included enslaved people, a reminder that “who worked the fields” was not only an economic question.
The twist is in what did not happen. The work did not vanish because people ate less. It vanished because each farmer, with machines, fertilizer and better seeds, produces far more. The next two episodes follow that story from the plough to the tractor.
Time travel
From the field to the office in two centuries
In 1800, about three in four American workers farmed. Today it is about one and a half in a hundred, and half the world's workers are in services.
United States: share of workers in farming
1800–1960: labor force aged 10+ (Lebergott); 2025: employed persons (ILO). Different concepts, same direction.
World: where people work
Share of total employment, ILO modelled estimates
Sources: US Census Bureau, Historical Statistics of the United States, Colonial Times to 1970, Series D 167–181 (Lebergott); World Bank WDI, SL.AGR/IND/SRV.EMPL.ZS (modelled ILO estimates, updated 8 Oct 2026). Shares calculated from the published counts.
Powered by Thorsten Meyer AI
The sector clock: Japan, China, Germany
To follow these shifts through the season, we use one recurring board: the sector clock. It shows the share of workers in each box for one country and one moment in time.
Japan is a clean case, because its census reports exactly these three sectors. In 1920, 53.8% of workers were in the primary sector, 20.5% in the secondary and 23.7% in the tertiary. In 1950 the field still held 48.5%. Then came the industrial boom: by 1970 the split was 19.3%, 34.0% and 46.6%. In the 2020 census it was 3.2%, 23.4% and 73.4%.
The sector clock
Japan: from field to office in a century
Share of employed persons by sector in Japan's population census.
- Field53.8%
- Factory20.5%
- Services23.7%
Gainful workers, all ages; about 2% not classifiable
- Field48.5%
- Factory21.8%
- Services29.6%
Employed persons aged 14+
- Field19.3%
- Factory34.0%
- Services46.6%
Employed persons aged 15+
- Field3.2%
- Factory23.4%
- Services73.4%
Unknowns imputed
Definitions changed over time: age limits and coverage differ by census year.
Sources: Statistics Bureau of Japan, 2005 Population Census, Population of Japan, Table 5.3 (1920–1970); Statistical Handbook of Japan 2025, Table 3.2 (2020). Mining counts as secondary, utilities as tertiary.
Powered by Thorsten Meyer AI
China moved even faster. Its statistics office has reported employment by “three strata of industry” since 1952. In 1952, 83.5% of the employed worked in the primary sector. In 1978, the year the reforms began, it was 70.5%. In 2000 it was exactly half. By 2024 it was 22.2%, while services had grown to 48.8% of all employment. Services overtook industry in 1994 and overtook farming in 2011.
The sector clock
China: the fastest move off the land
Share of employed persons at year-end by “three strata of industry”.
- Field83.5%
- Factory7.4%
- Services9.1%
- Field70.5%
- Factory17.3%
- Services12.2%
Reforms begin
- Field50.0%
- Factory22.5%
- Services27.5%
- Field22.2%
- Factory29.0%
- Services48.8%
Latest year published
Series break around 1990 (data revised after the 1990 census).
Source: National Bureau of Statistics of China, China Statistical Yearbook 2025, Table 4-2. Services overtook industry in 1994 and farming in 2011.
Powered by Thorsten Meyer AI
Germany followed Clark’s textbook pattern. In West Germany in 1955, industry was the biggest box, with 47.1% of employment, against 18.5% in farming and 34.4% in services. Industry peaked at 49.2% in 1965. In 1972, services overtook it for good. In 2025, according to the federal statistics office Destatis, 1.2% of people working in Germany were in the primary sector, 22.9% in the secondary and 75.9% in services.
A word of caution about all of these series: they use different concepts (labor force or employed persons), different age limits, and different rules for where mining, utilities or repair work belong. Japan’s early censuses count “gainful workers” of all ages; China’s series has a break around 1990; Germany’s figures cover only West Germany until 1990. The levels are not perfectly comparable across countries. The direction, though, is the same everywhere.
The number to remember: 2001
For most of human history, the most common kind of work was working the land. According to the ILO’s modelled estimates, published through the World Bank, that changed only very recently. In 1991, 42.8% of the world’s workers were in agriculture, 21.8% in industry and 35.4% in services. Services edged past farming in 1999, fell back in 2000, and have been ahead since 2001 (40.2% against 39.0%). In 2025 the split is 25.8% agriculture, 23.7% industry and 50.6% services.
So the world became a “service economy” about 25 years ago. These are modelled estimates, which the ILO revises as new surveys come in; the exact year could shift by one or two, but the turn itself is clear.
Myth check: “America doesn’t make things anymore”
Verdict: false. US factory employment did fall hard. In the Bureau of Labor Statistics’ payroll series, manufacturing jobs peaked in June 1979 at 19.55 million. In September 2026 they were 12.65 million, about 35% fewer.
But output tells a different story. The Federal Reserve’s index of real manufacturing output averaged about 1.9 times its 1979 level in 2025. America makes a lot; it needs far fewer people to do it.
One honest caveat: the same index peaked in December 2007 and has stayed below that peak since. US factories produce much more than in 1979, but not more than in 2007. “Fewer hands, more output” is the long-run pattern, not a promise of endless growth.
A fourth box, and a fifth?
Three boxes soon felt too few. In 1953, the American sociologists Nelson Foote and Paul Hatt, analysing what they called Clark’s “miscellaneous category” of tertiary industries, split it up. They kept domestic and personal services as tertiary and proposed a quaternary sector (transport, commerce, communication, finance and administration) and a quinary sector (medical care, education, research and recreation, including the arts). In 1961, the geographer Jean Gottmann described a “quaternary family” of work built on transactions, analysis, research and decision-making.
The definitions never settled. Today, some textbooks use “quaternary” for knowledge and information work and “quinary” for top decision-makers, a convention we could not trace to a single original source. Later episodes (8 and 41) come back to this, because the question of a new box is exactly the question of this season.
The season question: will automation become a sector of its own?
Machines have always moved work between the boxes. What is new is how much of the work machines might do on their own. In 2025, the International Federation of Robotics counted 5 million industrial robots operating in factories worldwide, more than twice as many as seven years earlier. In 2024, South Korea had 1,220 robots per 10,000 manufacturing employees, Germany 449, Japan 446, the United States 307 and China 166 (IFR figures; China’s density was recalculated with new employment data).
So here is my thesis, and I want to label it clearly as a thesis, not a fact: within a decade, or a few decades, automation could become a sector of its own, work done with no human hand, up to 90% of it. If that happens, the sector clock would need a fourth or fifth color that is not a group of people at all.
Fourastié would disagree. So far, every time machines took over the work in one box, people moved into the next one, and the fear of lasting technological unemployment proved wrong. The evidence for both views will build up over the season. In Episode 48, I put my thesis on trial, with the evidence for it and against it.
What this means for businesses
- Shifts move work; so far they have not ended it. Every big sector shift so far moved jobs to new places. The companies that did well were the ones that saw where demand was going next, not the ones that defended the old box longest.
- Map your own tasks. Sort the work in your company into field, factory and office tasks, and ask which of them a machine could do now, which in five years, and which depend on trust, judgement or physical presence.
- Watch productivity, not only headcount. The factory story shows that output can rise while jobs fall. Measure what your people produce, not just how many there are.
What this means for you
- See your job in the boxes. Most people in rich countries now work in services, the box that Fourastié thought machines could hardly touch.
- That box is now in play. AI reaches into offices, customer service and analysis, the work that used to be the safe destination. That is a reason to pay attention, not to panic.
- Follow the season. Over the next 59 days we look at each sector, each wave of automation, and what history says about the people who moved. General information, not career or financial advice.
What the numbers can’t tell us
- Definitions change. Sector boundaries differ between countries and over time; early figures are estimates.
- Shares are not levels. A falling share of farm workers can mean more food, not less; a falling share of factory jobs can come with rising output.
- Modelled estimates are revised. The ILO figures for recent years are projections and will be updated.
- History is not a law. That workers always found new work in the past does not prove they always will. That is precisely the question this season asks.
Series navigation
- Previous episode: this is Episode 1.
- Next episode (tomorrow): Episode 2, “When Almost Everyone Farmed“: what a working day looked like when three in four Americans worked the land, and how so few farmers ended up feeding everyone.
- Related on Thorsten Meyer AI: What Is Work For? A survey of 8,234 workers on what we lose if AI takes the job · What happens after an AI layoff? Goldman Sachs sees the same scars as 40 years of tech shocks
Sources
- Allan G. B. Fisher, The Clash of Progress and Security, Macmillan, London, 1935; “Production, Primary, Secondary and Tertiary”, Economic Record 15(1), June 1939, pp. 24–38; Economic Progress and Social Security, Macmillan, London, 1945, pp. 5–6 (quotes). archive.org
- Colin Clark, The Conditions of Economic Progress, Macmillan, London, 1940, pp. 6–7, 176, 182. archive.org
- Colin Clark, 3rd edition (1957), p. 490, as quoted in Zoltan Kenessey, “The Primary, Secondary, Tertiary and Quaternary Sectors of the Economy”, Review of Income and Wealth 33(4), 1987.
- Jean Fourastié, Le Grand Espoir du XXe siècle, Presses Universitaires de France, Paris, 1949; review signed A. S., Population 4(2), 1949, p. 380. persee.fr; Gallimard “Tel” edition 1989, publisher’s presentation.
- Nelson N. Foote and Paul K. Hatt, “Social Mobility and Economic Advancement”, American Economic Review 43(2), May 1953, pp. 364–378. archive.org
- Jean Gottmann, Megalopolis, Twentieth Century Fund, New York, 1961, pp. 576–577. archive.org (MIT Press open access)
- US Census Bureau, Historical Statistics of the United States, Colonial Times to 1970, Part 1, Series D 167–181 (Lebergott), p. 139. census.gov
- US Bureau of Labor Statistics, Employment Projections, Table 2.1 “Employment by major industry sector”, 2025. bls.gov
- US Bureau of Labor Statistics, Current Employment Statistics, series CES3000000001 (manufacturing employees). bls.gov
- Federal Reserve Board, G.17 Industrial Production and Capacity Utilization, manufacturing (SIC) index B00004, release of 16 October 2026. federalreserve.gov
- World Bank, World Development Indicators: employment in agriculture, industry and services (% of total employment, modelled ILO estimate), SL.AGR.EMPL.ZS, SL.IND.EMPL.ZS, SL.SRV.EMPL.ZS, last updated 8 October 2026. worldbank.org
- Statistics Bureau of Japan, 2005 Population Census, Population of Japan, Table 5.3 (1920–2005); Statistical Handbook of Japan 2025, Table 3.2 (2020). stat.go.jp
- National Bureau of Statistics of China, China Statistical Yearbook 2025, Table 4-2 “Number of Employed Persons at Year-end by Three Strata of Industry”. stats.gov.cn
- Statistisches Bundesamt (Destatis), “Erwerbstätige im Inland nach Wirtschaftssektoren” (long series). destatis.de
- International Federation of Robotics, “Five Million Robots now Operate in Factories Globally”, 24 September 2026, and “Robot Density Surges in Europe, Asia, and Americas”, 8 April 2026. ifr.org
This article is for information only. It is not career, legal, financial or investment advice. Historical figures are estimates as published by their sources; modelled estimates may be revised. Thorsten’s thesis on automation is an opinion, labelled as such, and is tested in Episode 48.
Researched and drafted with AI assistance, including Anthropic’s Claude.
