Has Technology Ever Reduced Labor?

Has technology ever reduced labor? The question sounds rhetorical. We carry small computers that answer any factual query in seconds, our laundry tumbles itself clean while we sleep, our cars drive themselves on highways our great-grandparents traveled by mule. Of course technology has reduced labor. The question barely needs asking.

The honest answer is no. Not in any aggregate sense, and not for any sustained period across the twelve thousand years of recorded technological development. Technology has reduced specific labors and replaced them with different labors, almost always under tighter supervisory discipline and almost always at higher total volume. The case that technology reduces work survives only by cherry-picking individual tasks, the washing of one shirt, the plowing of one acre, the copying of one manuscript, and ignoring where the freed time gets reallocated.

This is not a comfortable claim and the evidence for it is older and stronger than most readers expect.

The Foraging Baseline

The anthropological record makes the picture clearest. Marshall Sahlins published “The Original Affluent Society” in 1972 after collecting field data on hunter-gatherer time use from researchers working with the Hadza in Tanzania, the !Kung San in the Kalahari, and several Australian Aboriginal communities. The numbers were striking enough that Sahlins felt the need to defend them against disbelief from his own profession. Subsistence labor among these groups ran to roughly fifteen to twenty hours per week. The remaining waking time was given to storytelling, kinship work, ritual, sleep, conversation, and play.

The Neolithic transition to settled agriculture lengthened the workday while shortening life expectancy. Skeletal remains from early farming sites in the Levant, the Yangtze valley, and Mesoamerica show shorter average stature, more arthritis, more dental caries, and heavier parasitic loads than the foraging populations they replaced. Tooth enamel records seasonal hunger that does not appear in pre-agricultural skeletons. The first technological revolution in human history, the domestication of cereals and animals, made daily labor heavier than what came before.

The industrial revolution made it heavier again. Twelve and fourteen hour factory shifts replaced the seasonal agricultural rhythms that had at least permitted fallow winters, festival weeks, and the unstructured time between harvest and planting. E. P. Thompson documented this transformation in “Time, Work-Discipline, and Industrial Capitalism” (1967), tracking how factory clocks dismantled the older task-oriented sense of time and replaced it with the timed shift, the punched card, the docked wage. Each technological wave expanded productive capacity, and the working day expanded with it.

The Appliance Paradox

The most rigorous study of household technology is Ruth Schwartz Cowan’s “More Work for Mother,” published in 1983. Cowan tracked time-use diaries from American households across the twentieth century, from 1900 through the 1970s. Her data covered the arrival, in sequence, of the vacuum cleaner, the electric iron, the refrigerator, the washing machine, the dishwasher, and the clothes dryer.

Total hours spent on housework by middle-class American women held flat across the entire period at roughly fifty per week. Standards rose to consume every minute the appliances released. Floors that had been swept weekly were now expected to be vacuumed daily. Bed linens that had been laundered monthly were now expected to be changed weekly. Clothing that had been worn for several days between washings was now expected to change each morning. The wringer washtub vanished and was replaced by an enlarged regime of domestic cleanliness whose maintenance absorbed the savings entirely.

Effective because the appliances genuinely automated specific operations and reduced the muscular strain of laundry, dishwashing, and floor care. Not effective because the cultural definition of an acceptable household expanded to fill the available capacity, and because servants disappeared from middle-class homes during the same decades, transferring labor that had been distributed among several women onto the shoulders of one. The middle-class housewife of 1970 worked the hours of the 1900 housewife with better tools and no help.

Jevons and His Coal

The structural principle behind these findings was articulated more than a century before Cowan wrote. William Stanley Jevons published “The Coal Question” in 1865 to address what seemed an obvious puzzle. Steam engines had become dramatically more efficient over the preceding fifty years. Each ton of coal now produced more useful work than ever before. By the logic that informs most contemporary discussions of technology, British coal consumption should therefore have fallen.

It had multiplied. Cheaper steam power made new applications economically viable, opened new markets, lowered the cost of goods and so expanded their demand, and pulled coal into uses no one had imagined when the engines burned twice the fuel for half the output. Jevons concluded that efficiency gains in any input commodity tend to increase rather than decrease its total consumption.

The same dynamic governs labor. Email was sold to office workers in the 1990s as a time-saving substitute for paper correspondence and inter-office memos. A McKinsey Global Institute study in 2012 estimated that knowledge workers now spend twenty-eight percent of the workweek processing email, a category of activity that barely existed in 1985. The smartphone, sold as a tool to make office work portable for the executives who needed it, dissolved the boundary between work and not-work for everyone who carried one, with the consequence that labor now extends into bedrooms, vacations, and predawn hours. Slack, Microsoft Teams, and the productivity software stack of the 2020s have multiplied this effect.

What Capital Selects For

Harry Braverman’s “Labor and Monopoly Capital,” published in 1974, supplied the structural argument that ties the historical pattern together. Under capitalism, technology is selected and deployed to increase surplus extraction. That goal usually requires deskilling work, so that cheaper and more interchangeable workers can replace expensive skilled ones, and accelerating pace, so that more output emerges per labor hour purchased. Reducing aggregate labor is incidental to this goal at best and adversarial to it at worst. The Ford assembly line of 1913 produced eight times more Model T chassis per shift at the same daily wage and the same daily hours. Henry Ford paid his workers to produce more cars in the same time, never to produce the same cars in less time.

David Graeber carried the argument forward in “Bullshit Jobs” (2018). John Maynard Keynes had predicted in his 1930 essay “Economic Possibilities for Our Grandchildren” that the productivity gains of the coming century would deliver a fifteen-hour workweek by 2030. The technical capacity arrived on schedule. Productivity per worker hour in the United States grew by roughly two hundred and forty percent between 1948 and 2024 according to Bureau of Labor Statistics figures. American workers in 2026 still average roughly thirty-eight to forty-two hours per week, and white-collar workers exceed forty once unpaid evening and weekend labor is counted into the total. The productivity gains were absorbed by an expansion of administrative, supervisory, compliance, and ceremonial labor whose function is to manage other workers or perform the appearance of work rather than to produce anything an earlier generation would recognize as output.

The Agricultural Counterexample

A careful reader will object. A wheat farmer in 1850 worked harder physically than a Kansas farmer in 2026 operating a GPS-guided combine with a climate-controlled cab and satellite yield monitoring. This is true and not in dispute.

The question concerns labor in the aggregate. Two percent of the American population now farms. In 1850, seventy percent did. The displaced sixty-eight percent did not retire to the leisure their grandparents had imagined. They went into factories, then offices, then call centers, then warehouses, then gig-platform delivery routes, then content moderation queues for social media companies. They were typically subjected to closer time-discipline in their new occupations than they had experienced on the farm, where the boss was the sun and the seasons. The combine reduced agricultural labor while the displaced workers found that their total work simply migrated into other sectors at unchanged or higher intensity.

Total hours worked by American workers fell sharply between 1900 and 1950, from a manufacturing average near sixty per week to a postwar average near forty. That reduction was won by organized labor through strikes, legislation, and collective bargaining, with technology playing the role of the productive base that made shorter hours economically survivable for employers rather than the role of the cause. Productivity has since more than tripled, and the workweek has barely moved. The single sustained reduction in American work hours in modern history coincided with the rise of unions, and the flatlining of hours after 1950 maps directly onto the decline of union density. The arc since the digging stick has bent toward more work, with one twentieth-century reversal that political organization carved out and that capital has been clawing back ever since through gig contracts, salaried-exempt classifications, and the always-on mobile workforce.

The Real Exceptions

There are genuine exceptions worth naming, because they reveal the conditions under which technology actually does reduce labor.

Anesthesia eliminated the labor of restraining a screaming patient during amputation. Antibiotics eliminated weeks of bedside nursing for survivable infections like pneumonia and sepsis. The printing press eliminated the scriptorium copyist. The mechanical loom eliminated certain categories of hand-weaving without substitute. These are real cases of labor reduction, and they share a feature worth noticing.

Each eliminated a task whose demand could not be inflated. You cannot expand demand for amputation-restraint the way you can expand demand for clean floors. You cannot generate more bedside-nursing-for-untreated-pneumonia the way you can generate more email correspondence. The demand is bounded by the underlying biological or material reality, and once the technology meets that demand the labor disappears for good.

Where technology hits a hard ceiling on demand, it reduces labor. Where demand is elastic, and almost all labor categories have elastic demand, it does not. The contemporary knowledge economy is built almost entirely on elastic-demand labor. There is no natural ceiling on the number of meetings a corporation can hold, the number of reports it can file, the number of slide decks it can produce, the number of compliance forms it can require, or the number of performance reviews it can conduct. Each productivity tool that arrives in this sector arrives in an environment that will absorb its capacity without bound.

The Generative AI Frame

This is the relevant context for the present moment. Generative AI is being marketed to workers as a tool that will give them their evenings back. It is being deployed by employers to intensify and discipline existing labor through productivity dashboards, automated performance review, output quotas measured in tokens per hour, and surveillance systems that track keystroke cadence and screen attention. The same pattern Cowan documented for the dishwasher and Braverman documented for the assembly line is repeating in the knowledge sector in real time.

Early evidence already supports the structural prediction. Brynjolfsson, Li, and Raymond’s 2023 NBER study of customer service workers using generative AI assistance documented a fourteen percent average productivity increase, and follow-up reporting in 2024 and 2025 indicates that several large operators have responded by raising ticket throughput targets to absorb the gain. Anecdotal accounts from software engineers at firms deploying coding assistants describe rising line-of-code and feature-velocity expectations alongside the productivity tools, with stress reports trending upward rather than down. The pattern is repeating in the open, with confirmation arriving in the same quarterly earnings calls that promise the technology will save us all time.

Why This Matters

The premise embedded in the question, that technology should or could reduce labor, is a nineteenth-century progressive assumption the twentieth century falsified and the twenty-first is falsifying again. Labor is not reduced because no economic system in human history has been designed to reduce it.

Capitalism captures productivity gains as growth, profit, and expanded managerial overhead. State socialism in its twentieth-century forms captured them as expanded production targets and political mobilization, with Stakhanovite competitions celebrating workers who exceeded their quotas rather than workers who finished early. Subsistence economies captured productivity gains as expanded population, the Malthusian mechanism running its course before any surplus could accumulate. The only societies that ever worked materially less than the modern average were the foraging bands whose entire technological inventory was a digging stick, a sharpened rock, and the accumulated ecological knowledge of several hundred generations.

The arc since then has bent in the opposite direction. There is no current technology, including the one currently being marketed as the most labor-saving innovation in human history, that shows any sign of bending it back. The decisive variable has never been the technology itself. It has always been the question of who owns the productivity gain and what they are permitted to do with it. That is a political question, not a technical one, and no engineer working at any company anywhere is going to answer it for us.

The argument here connects to several other pieces in the BolesBlogs archive worth reading alongside it. A companion piece on the Rehearsal State examines how performative governance has replaced the working kind, which is the political twin of how performative work has replaced the productive kind. Another piece, the Rental Life article, tracks the same productivity-capture mechanism in the consumer sector, where goods themselves stop being owned and start being leased back to us by the month. On the cognitive side, the zeroing of knowledge by AI piece addresses what happens to the thinking labor that survives the productivity squeeze, which gets simulated, then standardized, then absorbed. Most directly relevant of all is Carceral Nation, the first volume of the Institutional Autopsy trilogy, which traces how the prison and the workplace converged on the same disciplinary architecture during the same century that the productivity gains documented above were being captured upward rather than redistributed downward.

Anyone who tells you otherwise is selling something, usually a subscription.