The Robot Apocalypse That Keeps Creating New Paperwork: One Hundred Years of Jobs Automation Was Definitely Going to Destroy

In 1930, John Maynard Keynes — a man who was right about enough things that we still teach him in universities — predicted that by 2030, the standard workweek would be fifteen hours. Automation would handle the drudgery. Humans would finally be free to pursue leisure, contemplation, and the finer things. He was off by about thirty-five hours per week, depending on your industry and how honestly you're counting your email time.

This is not a story about Keynes being wrong, exactly. It is a story about a prediction that has been made, in various forms, by very smart people for one hundred years — and what actually happened instead.

The Assembly Line Panic of the 1950s

The first modern wave of automation anxiety arrived in the postwar boom, when American factories began introducing mechanized assembly processes at scale. In 1955, a New York Times headline asked whether automation would produce permanent unemployment. Walter Reuther, the influential head of the United Auto Workers, testified before Congress that machines were replacing workers faster than the economy could absorb them.

He was not wrong that displacement was happening. He was wrong about what came next. The economy absorbed the workers — into new industries, new roles, and entirely new job categories that hadn't existed a decade earlier. The television repairman. The appliance salesman. The suburban shopping center clerk. Automation eliminated certain repetitive factory tasks and, somewhat inconveniently for the prophecy, helped generate enough economic growth to create millions of jobs doing other things.

The robots had arrived. The unemployment apocalypse had not. Reformers recalibrated and waited for the next wave.

The Computer Is Going to Take Your Job (1970s–1990s)

The personal computer gave the automation-anxiety industry a genuine second wind. By the early 1980s, serious economists were publishing papers suggesting that white-collar work was about to be as vulnerable as blue-collar work had been in the 1950s. Word processors would eliminate secretaries. Accounting software would eliminate bookkeepers. Spreadsheets would eliminate whole floors of number-crunchers.

Some of this happened. The typing pool did largely disappear. But here is where the prophecy kept tripping over the same rock: automation eliminated tasks, not jobs, and humans are remarkably creative about filling the time that gets freed up with new tasks that require new jobs.

The spreadsheet didn't eliminate financial analysis — it made financial analysis so fast and cheap that companies started doing vastly more of it, employing more analysts than before. The word processor didn't eliminate office workers — it enabled a proliferation of documentation, communication, and administrative complexity that actually increased the demand for office workers. This phenomenon, which economists would eventually call the "productivity paradox" and related effects, was not part of the original forecast.

The 2013 Oxford Study and Peak Robot Panic

If the automation-anxiety industry has a single greatest hit, it is probably the 2013 Oxford study by Carl Benedikt Frey and Michael Osborne, which concluded that approximately 47 percent of American jobs were at high risk of being automated within the next decade or two. The paper was cited everywhere. It was the academic equivalent of a viral tweet. Nearly half of all jobs. Journalists wept. Think tanks convened. Presidential candidates mentioned it in debates.

A decade later, U.S. unemployment was at historic lows. The jobs that were supposed to have vanished were, in many cases, still there — sometimes transformed, sometimes augmented by the very technologies that were supposed to eliminate them. Radiologists, whom the study flagged as highly automatable, were still employed in large numbers, now often working alongside AI tools that handled preliminary image screening and made them more productive rather than redundant.

Frey and Osborne's methodology was subsequently critiqued by other economists who pointed out that the study analyzed tasks and extrapolated to jobs in ways that overstated the replacement risk. The 47 percent figure, it turned out, was more of a conversation starter than a prophecy. A very, very effective conversation starter.

The AI Moment (We Are Currently In It)

The current chapter of this story involves large language models, generative AI, and a fresh cohort of forecasters explaining that this time the displacement will be different, faster, and more comprehensive than anything before. Some of these forecasters are the same people who made similar claims about the internet, about mobile computing, and about the 2013 automation wave.

They may be right that something significant is happening. The technology is genuinely remarkable. But the track record of "this wipes out X percent of jobs" predictions is, to put it gently, not strong. What the last hundred years suggest is that automation tends to eliminate specific tasks, shift the composition of work, and generate demand for new roles — including, delightfully, roles specifically dedicated to overseeing, auditing, explaining, and occasionally apologizing for the automated systems.

The AI era has already produced AI prompt engineers, AI ethics officers, AI output reviewers, and an entire consulting industry devoted to helping companies figure out what to do with AI. The robots came for the jobs. The jobs mutated. We hired people to supervise the robots.

The Consistent Error

What every generation of automation prophets has underestimated is the same thing: human beings' peculiar talent for finding new things to do. We did not stop working when the washing machine eliminated hand-laundering. We did not stop working when the tractor eliminated most agricultural labor. We invented suburbs, service industries, knowledge work, the gig economy, and, eventually, jobs with titles that would have been completely incomprehensible to anyone in 1950.

"Social Media Manager" would have required a lengthy explanation. "UX Researcher" would have produced blank stares. "Growth Hacker" would have resulted in a call to the authorities.

Keynes' fifteen-hour workweek is still out there, somewhere in the future, perpetually five technological revolutions away. The robots are getting better. The paperwork, somehow, is getting worse.