Don't Be Afraid: Each of These 30-Odd Jobs Has Been Sentenced to Death at Least Once — and Look, They're All Still Fine

New technology really has destroyed some jobs, but it has created others. And what happens in the workplace also turns on the economy, employment policy, labor supply and demand, and how work gets divided up.

Ever since humans invented the machine, there has probably been no such thing as a job for life. Research institutes and academic bodies seem to have made a pact among themselves: every so often they publish a few warnings about "robots replacing human labor," and each one touches off a fresh round of argument.

The harshest recent warning came from McKinsey. It scored 800 occupations and more than 2,000 work activities for automation risk and told people that yet more jobs were about to be automated away. Assembly-line workers in the factories of food giants were in the most danger, and 86% of a librarian's work would not escape either.

The argument comes back year after year, and the positions barely change. The optimists like to point to the cost of robots and the immaturity of the technology to prove that human labor is still better. The radicals carry on the spirit of the Luddites of 200 years ago and call robots "job stealers" outright — though since artificial intelligence isn't going to amount to much any time soon, most of them can't be bothered to make a fuss. The moderates sit in the middle, sounding especially politically correct.

Out of curiosity, we went and read through nearly 60 years of papers, reports, and press coverage on machines, robots, automation, and AI replacing human labor. We found dozens of jobs that had been declared "about to be wiped out completely" — some of them more than once. Here they all are.

Administrative clerks, secretaries, assistants

These jobs have existed as long as the modern company has.

It is an awkward line of work with little visible presence — send a few too many mass emails and people complain you have too much time on your hands — and it was pronounced dead 60 years ago.

Around 1950, when computers were first brought into office data work in the United States, experts began to warn that "most administrative work will be automated away" — in every field, not just companies but schools, public services, even government. The prediction has never really stopped since.

Administrative staff get a raw deal, honestly. They look as though they hold the power to keep an organization running, but they are more lubricant than engine — a support function, and they sometimes joke that they are "logistics." When results turn bad and layoffs loom, they are often the first to be considered.

If a cheaper substitute is available — a time clock instead of an attendance clerk, Siri instead of a secretary, scheduling software instead of an assistant — why not?

But there is no need to worry too much. Administrative jobs weathered the storm and are still doing fine. From 1957 to 1987, the number of administrative positions in the United States grew steadily and fast, becoming an employment group second in size only to managers, executives, and service workers.

How did the prediction get it so wrong?

Computers did slow the ordinary growth of administrative work. But they also cut the cost of tasks that used to be expensive in both time and money — now reports can be tallied and data analyzed every day. So computers let companies in every field build bigger businesses, and that created jobs instead.

Finance: bank tellers, cashiers, accountants, auditors

Here is another job most companies need but don't think of as important. In 1968, a report listed the 10 most common areas for industrial automation: payroll, accounts receivable, turnover, sales data analysis, general ledger management, time allocation, surveying, inventory control, expense calculation, and cost accounting — nearly all of them finance.

By the 1990s, as cash machines spread across American cities and suburbs, banks in the United States cut 179,000 teller jobs in a decade, 37% of the workforce. By 2000, it was projected, 90% of bank customers would use an ATM.

By 1995, Deloitte predicted that computerized office work would push banks on both sides of the Atlantic to cut 50% of their jobs. For the U.K., the figure was 50,000 people.

Twenty years on, the replacement forecasts for finance jobs are harsher still. Not long ago McKinsey reported that in occupations such as accounting and auditing, 86% of work activities carry automation risk.

Customer service: front desk, telephone operators, salesclerks, shopping guides

"For Mandarin, press 1. For English, press 2." If you don't go on to press 0 (or 9) for a human, whatever solves your problem next will be a machine on some other line.

The claim that customer-service jobs were doomed showed up as early as the 1980s, maybe earlier. Even before any of the clever systems existed — vending machines, TV shopping, cloud service platforms — product manuals and buying guides had been around a long time.

And customer service may look like an endless stream of small details, but customer needs really fall into only a few kinds:

Banks — report a lost card, check a balance, change a password. Malls — opening hours, promotions, whether a product is genuine. Carriers — top-ups, bill payment, where the nearest store is. As long as companies break every service situation down and keep adding to the list, these needs can all be sorted into categories and handled by reusing the same answer over and over.

But look: everyone shops on Taobao now, so why is there still a job called dianxiao'er, the shop attendant? You can report a lost bank card by phone, on the official website, in a mobile app, even by yourself on WeChat — so why do people still like to ask for a human?

It is like asking whether the watermelon is sweet every single time you buy one. You know the answer will probably always be the same, but hearing someone say it out loud is more reassuring, especially when something unexpected happens. That is why demand for human service has stayed stubbornly high.

Middle managers

This one may be the biggest surprise.

In some risk assessments, certain institutions do put corporate management among the jobs least likely to be automated away.

Yet in 2014, a Japanese venture capital firm named an artificial intelligence called Vital to its board, hoping to use its sharp feel for market trends.

Go back another 20 years and Fortune had a forecast of its own: the working conditions humans built 200 years ago — mass production and sprawling organizations — were disappearing. As labor redistributed, three-quarters of middle managers would go, some of them possibly back to the jobs they started in. After that only two kinds of leader would be left: process managers, who handle execution and oversight, and employee coaches, who handle development.

The result? As you can see, middle managers still count as the backbone of the company, at least for now.

Creatives: designers, journalists, professors, lawyers, architects

Predictions that these jobs would be replaced have mostly come in the past few years.

In 2014, for instance, a company called Narrative Science launched software that turns raw data straight into readable text, replacing reporters or analysts. Word has also gone around about robots writing novels.

But in Oxford University's 2013 ranking of more than 700 occupations by risk of computerization, creative jobs came out low. Lawyers scored just 0.0035, graphic designers 0.082, journalists 0.11.

McKinsey's recent automation-risk report bore that out. "Given technology available today, the hardest activities to automate are those involving managing and developing people (only 9% automation potential), and those requiring expertise applied to decision-making, planning, or creative work (18% potential)."

These jobs call on intelligence and feeling that only humans have, so there is almost no chance they will be permanently and completely taken over by automation or AI.

Craftspeople: blacksmiths, fishermen, cooks, bakers, bartenders

As Johnson joked in The Dead Beat, death always comes in a string. In the 1970s and '80s, anyone cheered by the rise of the car industry could reach the correct conclusion without much effort: blacksmiths and carriage-makers were about to leave the stage of history.

More than 30 years ago, jobs that depended on craft skill were already forecast to grow more and more slowly than the average.

Faced with a huge, standardized, high-efficiency, low-cost industrial and commercial system, the part craft plays in getting a piece of work done keeps shrinking. With purpose-built fishing boats and electric nets, why use a hook and a net? Once the dough mixer, the automatic recipe machine, and the timed oven were invented, baking bread no longer needed anyone watching over it.

But craftspeople go by another name now — artisans — even if "artisan spirit" has been worn out from overuse. And there is no stopping it: anything labeled hand-made is simply more appealing.

Middlemen

Look at the brokerage offices on every street corner today. Doesn't it make you want to laugh?

Then again, the internet industry may talk a smooth game, but it really has cut out the middleman in some cases.

As Erik Brynjolfsson and Andrew McAfee write in their book The Second Machine Age, who could have imagined that people with rooms to rent would connect directly with renters on a platform like Airbnb, taking over the role of hotels?

But the trend was predicted as early as 1972. That was when a device called a card separator was invented and people realized they could buy a plane ticket directly, without going through any middleman at all.

Later, as home computers spread, Reuters predicted in 1995 that middleman jobs in banking, insurance, and travel were very likely to be "destroyed."

Today, job seekers send résumés straight to company websites, people find apartments on local-services platforms, and even used-car platforms claim that individual owners meet individual buyers directly. But look: headhunting firms are still here, and Lianjia has all but taken over Shanghai.

Thomas Friedman, in The World Is Flat, described the ten forces that flattened the world, most of them tied to new technology. In a flatter world, the two ends of a commercial exchange have more and more chances to reach each other directly, which naturally narrows the room left for a middleman. And yet what happened is that another kind of middleman appeared.

Administrators: warehouse managers, sorters, librarians, mail carriers, security guards

Turn the clock back to the 1970s and you can already find plenty of articles on automating library management and on models for applying such systems.

Then in 1982, Japan stepped into the shift toward management-system technology. Toshiba spent two years and US$8 million building what may have been the world's most advanced mail-sorting system at the time. It weighed, scanned, calculated postage, and stamped mail automatically, then dropped it into the right delivery bin. The postal system still needed people back then — the machine could handle only 60% of the work — but that was a sign that 60% of the jobs were finished.

Thirty years on, the same thing is happening. The most famous case is Amazon, which bought the Kiva robotic system in 2012; Kiva robots now do the bulk of the sorting in its logistics warehouses. Chinese e-commerce platforms such as JD.com also bring up automated sorting whenever they talk about warehouse logistics.

In 2013, the American company Knightscope released K5, a crime-prevention robot that moves through a full 360 degrees, matches faces against a database of suspects, and can read license plates at night. The French company Abankos Robotic built an anti-theft robot, Lucos, and Cunos, a wolfdog robot that patrols on its own, to replace human security guards.

Drivers: chauffeurs, pilots

Would you feel safe in a vehicle with nobody driving it?

Boeing wanted to know the answer decades ago. A 1984 IFG report noted that Boeing had made a first attempt to replace one of the crew in the cockpit with a machine that would work alongside the other two, and predicted: "As computers get cheaper and cheaper, the chance of replacement will rise sharply even for a job as complex as pilot."

Now that autonomous driving is being hyped again, the discussion has resurfaced. But a string of recent self-driving crashes, above all the death of a Tesla owner, has put the idea back under a cloud.

In theory, that day may really come. But human drivers may never be replaced entirely.

Dangerous work: rescue workers, soldiers, astronauts, lab scientists

This work is hazardous and physically punishing, sometimes done in extreme conditions where humans may not cope, or cope only at high cost — which is exactly why people want robots to take it over. Scientists and engineers have been working on it for decades.

One unusual case: the 2001 anthrax attacks in the United States. Someone mailed letters containing anthrax bacteria to several news offices and two Democratic senators, and the postal sorters who handled them died or fell ill. Robots were then brought into mail-sorting systems, and in the end workers were kept safe.

In 2011, President Obama announced the "National Robotics Initiative", promising US$70 million for a next generation of robots to help factory workers, health care providers, soldiers, doctors, and astronauts do difficult work.

Attendants: supermarket clerks, cashiers, order takers, cleaners, gas station attendants

Decades have passed and the robot waiter has got no further than carrying a plate or shaving noodles. But some machines far more useful than robots became common long ago.

Take the barcode scanner. In 1982, Columbia University predicted that within two years half of the country's 23,000 large supermarkets would have scanners installed, and the other half would take another six years. In other words, by 1988, 50,000 supermarket jobs would be gone.

Then came debit cards and shopping by telephone. In 1984, researchers at Stanford University saw spending going cashless and concluded that cashier jobs would soon start to disappear.

And just as coachmen had already given way to motor vehicles, it would not be long before mail carriers and hotel staff gave way to automated systems.

Fifteen years later, McDonald's put touchscreen ordering machines, which customers liked, into some of its restaurants. Some concluded from this that the first wave of robots replacing human labor would hit fast food around 2030, then spread to hotels, malls, and airports, replacing most receptionists and cleaners.

Medical workers: pharmacists, surgical assistants, lab assistants, elder-care workers

As long as 30 years ago there were predictions that "smart" robots would replace doctors.

In 1996, a popular-science paper on robots in medicine raised the ways they might take over from people: lab robots to assist with experiments; hospital robots to move around, pick things up, and dispense drugs; rehabilitation robots to carry patients; surgical robots to make sure a procedure is precise.

In 2008, the U.S. National Intelligence Council predicted that by 2025 robots might be working as elder-care aides.

In recent years pharmacists, along with medical secretaries and medical transcriptionists, have also been named as jobs robots will gradually take over.

But these people are still busy in hospitals. And in the United States you will find that this kind of labor problem usually gets solved through immigration: as long as the labor market supplies enough people cheaply enough, employers still choose people.

Assembly-line workers, cleaners, print workers

Don't want to be a cog in the machine? You may not get the chance much longer.

In 1961, General Motors put the world's first industrial robot to work on the line at its New Jersey plant. Owners suddenly saw that the thing was fast and accurate, took no holidays, and would never go on strike. It was like a door opening onto a new world of capitalism.

So everyone copied it. In the 1960s, Ford's workforce fell from 65,000 to 35,000 even as the company built a string of new automated plants within a 100-mile radius.

The Society of Manufacturing Engineers began watching this new automation trend. In a 1980 survey, the society all but poured cold water on factory jobs with the conclusion that "robots will replace 20% of factory work within five years, and 50% within fifteen."

The next year, an analysis called The Impact of Industrial Robots stretched the timetable a little, arguing that over the next 20 years simpler first-generation robots would replace 1 million machine operators and more advanced second-generation robots could replace 3 million — at a time when the whole United States had only 8 million factory workers.

Because these were also depression years, and machines were advancing, the number of factory workers and private cleaners fell sharply in a single year, well below the levels of the years before.

By 1987, the printing industry was giving the working class no peace either. Electronic publishing replaced the old craft outright, and the number of printing and typesetting jobs fell year after year — it was less that machines replaced people than that the moment the machines appeared, the jobs simply ceased to exist.

Farmers

Like "worker," this is a way of naming an occupation, and it may not count as a job in the strict sense.

The risk forecasts for both arrived with the Industrial Revolution.

In 1885, the steam engine improved by the Briton James Watt lifted the lid on a change in how things were made. Not long after the machines arrived, the experts were proved right: the work got easier for a few years, then the work was gone. In the less than 150 years after 1850, agriculture's share of the American labor force fell from 60% to 2.7%.

Population change had something to do with it, but so did this: in the farming of that era, one machine was widely held to do the work of at least four farmhands, and if you ran it around the clock you could earn back what you paid for it within a year. Only a fool would not be tempted.

Then it got more complicated. People found that machines were not only cheap but could do, and do well, work humans were bad at. A 1970 report noted that farm laborers, cheap as they were in the short run, still could not match small machines that sowed and plowed more precisely — and it all but declared farming finished as an occupation.

But look: most of these jobs that were sentenced to death are still doing fine.

It should not be denied that robots and automation really have stolen some human jobs. In the United States, nearly 5 million manufacturing jobs have disappeared since 2000 — 30% of the total — and wages did just as badly, their share of national economic output falling to its lowest point of the 20th century.

But automation should not take all the blame, nor does this mean robots are so formidable. The economy, employment policy, labor supply and demand, how work is divided up, even a worker's own skills may matter as much to employment as any technological trend.

Many jobs disappear not purely because of new technology or the arrival of a machine, but as a natural result of the times moving on. Professional letter-writers vanished in China not only because the telephone, the mobile phone, and the internet were invented, but because education improved and more people could read and write.

And technology may have destroyed some old jobs, but it created new ones. A few years ago employers were hiring typists and switchboard operators; today the same employers may be hiring internet operations staff or mobile-game designers, at higher pay. Online shopping hurt the business of shop-floor sales staff, but it created a job called online customer service — and in China most online service agents are not chatbots. It created courier work too.

As Jason Furman, chairman of the U.S. Council of Economic Advisers, said in one report, you have to look at the economy over the long run: robots may reshape human jobs in the short and medium term, but within a period of economic equilibrium, AI's effect on the economy is still positive.

Which is to say that as technology advances, the world's productivity rises overall and its capacity to create wealth keeps growing. The simplest observation: the number of people the world can feed only goes up.

So don't be afraid.

Cover image from Sky and collider