Every generation is convinced its version of automation is the one that finally does it, the one that makes human labor optional. And every single time, we've been wrong about how it plays out. Not about the disruption itself, that part's real. We just keep misjudging the mechanism.
Here's the timeline, era by era, of technologies that were supposed to end human work, and what actually happened instead.
The Cycle We Keep Repeating
Gartner has a name for this: the Hype Cycle. A technology shows up (Innovation Trigger), gets wildly overhyped (Peak of Inflated Expectations), crashes back to earth (Trough of Disillusionment), then slowly earns its real place (Slope of Enlightenment, Plateau of Productivity).
Here's the thing that trips people up though: the technologies that actually stick around don't replace people. They change what people do. The ones that flop usually fail for the same reason: they misunderstood what the job actually involved in the first place.
The Mechanical Age (1811 to 1900)
The Luddites and the power loom. The story most people know is wrong. The Luddites weren't techno-phobic weavers scared of machines. They were skilled workers who understood the machinery perfectly well and objected to how factory owners were using it: to strip skill out of the job and replace them with cheaper child labor. The machines didn't wipe out textile work. They moved it into factories. Cottage industry died. The workers didn't.
Same story with the mechanical reaper and the typewriter. Each one was supposed to gut a whole category of labor. Instead they grew the economy so much that entirely new categories of work appeared.
The Early Computer Age (1950s to 1970s)
Before electronic computers existed, "computer" was a job title. Usually held by women with serious math backgrounds who did complex calculations by hand. When electronic computers arrived, those same people didn't disappear. Many of them became the first programmers and systems analysts.
Then came the first big AI promise, and the first big AI faceplant. In 1965, Nobel laureate Herbert Simon confidently predicted that machines would be able to do "any work a man can do" within twenty years. By 1973, a British government report concluded AI research had achieved almost none of its stated goals. Funding dried up almost overnight. That's the first AI winter.
The Software Era (1980s to 1990s)
Expert systems were supposed to replace doctors, lawyers, and engineers by encoding their judgment into if-then rules. They collapsed the moment they hit a situation nobody had thought to program for, because every single rule had to be written by hand. The maintenance load became unbearable. That's AI winter number two.
CASE tools and 4GLs promised that anyone could build software through diagrams and drag-and-drop modeling, no programmers required. What actually happened: the easy stuff got easy, and everything else still needed real engineers. Complexity didn't go away, it just moved.
Outsourcing was supposed to make domestic knowledge workers obsolete. By 2005, a quarter of major companies had already quietly brought those functions back in-house. Dell is the classic example, they reversed their offshored support after customers got fed up with reps who couldn't go off-script. Outsourcing didn't vanish, it just turned into something closer to staff augmentation.
The Automation Boom (2000s to 2010s)
ATMs vs. bank tellers. This is the one everyone cites, and everyone gets backwards. Teller employment actually went up for years after ATMs rolled out, because the cost savings let banks open more branches. The real decline came later, and it was digital banking as a whole, not the ATM by itself.
Self-checkout cut cashier jobs, but slower than predicted, and it created new roles in maintenance, loss prevention, and the whole discipline of designing customer experience.
Self-driving cars were supposed to be everywhere by now. Elon Musk said a Tesla would drive itself cross-country with zero human input by 2017. Ford promised steering-wheel-free Level 4 vehicles. Instead, Cruise shut down, Ford and Volkswagen pulled back their programs, and Waymo runs robotaxis in a handful of cities under tightly controlled conditions. Full autonomy is still boxed into small geofenced areas.
3D printing was going to put a printer in every home and end mass manufacturing as we knew it. It turned out to be genuinely great for aerospace parts, medical implants, and prototyping. Your local factory is unchanged.
The Cognitive Automation Wave (2015 to 2023)
RPA bots were pitched as capable of automating 40 to 70 percent of knowledge work. What companies actually used them for was handling growth without hiring, not firing people outright. In finance, headcount dropped mostly through attrition, not layoffs. In healthcare, it freed up clinicians to spend more time with patients.
No-code and low-code platforms promised that "anyone can be a developer" and professional engineers would become unnecessary. Instead, low-code became another tool in developers' kits. IDC actually projected the low-code developer population would grow more than three times faster than traditional developers, meaning more developers overall, not fewer.
Blockchain and smart contracts were supposed to make lawyers and banks irrelevant through trustless, automatic execution. Then $63 million got frozen in a buggy smart contract spanning three legal jurisdictions, and untangling it took armies of lawyers. The "trustless" system needed more human oversight than the traditional one it was replacing.
Chatbots and voice assistants were going to end call centers. They handle the simple stuff fine. Anything emotionally complicated or brand-sensitive still gets routed to a person. Projections have customer service roles declining a modest 5 percent by 2033, real, but nowhere near extinction.
The Generative AI Moment (2022 to now)
Now it's large language models, and the claim is the same one we've heard six times before: this will replace writers, artists, developers, lawyers, doctors, analysts.
So far, what we're actually seeing is augmentation. AI handles the routine, repetitive layer, people handle judgment calls and edge cases. There's also a quieter phenomenon worth noticing: once a technology becomes common enough and useful enough, people stop calling it "AI" at all. Spell-check, recommendation engines, voice-to-text, all of that was AI once. Now it's just software nobody thinks twice about.
And a reality check worth keeping in mind: a striking number of European startups that market themselves as "AI companies," roughly four in ten by some counts, don't actually use AI in any meaningful way.
What 200 Years Actually Teaches Us
We're always wrong about the mechanism, never about the disruption. Simon in 1965. Expert systems in the 80s. RPA vendors in 2017. LLM boosters today. Each one was sure they'd crossed the line into human equivalence. The line keeps moving.
Jobs are bundles of tasks, not single things. Technology doesn't erase people, it strips out specific tasks. ATMs didn't kill the teller job, they killed the cash-handling part of it and opened up the advisory part. RPA doesn't kill analysts, it kills data entry and creates more room for interpretation.
The Luddites were never anti-technology. They were anti-exploitation. Historian Brian Merchant's research is clear on this: they opposed using machines to de-skill workers and funnel profits upward, not the machines themselves. The actual danger was never replacement. It was degradation of the work itself.
Complexity doesn't disappear, it just moves somewhere else. CASE tools shifted complexity from code into diagrams. Low-code shifted it from syntax into logic design. AI is shifting it from execution into prompting and verification. Someone still has to hold the complexity. It just changes shape.
Nobody can see the new jobs coming. In 1900, a third of American workers farmed for a living. Mechanization gutted that number, and in its place came manufacturing, healthcare, education, and tech sectors that would have been unimaginable to a farmer at the time.
What This Actually Means
If you run a company, the useful question isn't "how many people can we cut." It's "what higher-value work does this free people up to do." The businesses that won the last several waves were the ones that redesigned roles, not the ones that just deleted them.
If you're an employee, the real threat isn't a machine doing your entire job. It's a machine doing 40 percent of it, which forces you to compete on price for whatever's left. The way out is the same as it's always been: keep moving your skills up the value chain.
If you're an investor, "replacement" is a great pitch. "Augmentation" is what actually builds a company that survives ten years.
The question was never really whether AI is going to replace you. It's whether you're going to be the one who decides what comes after the part of your job it takes over.
No comments:
Post a Comment