So let's do what these pieces almost never do: finish the sentence.
Job losses are real — that's not in dispute, and pretending otherwise would be its own kind of lie. But "AI will destroy jobs" is the first half of a sentence whose second half has been true every single time we've run this experiment for two hundred years. Real numbers, real sources, and the honest hard part included. Let's go.
---
## First, the part that's true: jobs are going
Start here, because softening it would be dishonest. The displacement is not hypothetical and it's not all in the future.
The World Economic Forum surveyed more than 1,000 employers representing over 14 million workers across 55 economies for its **Future of Jobs Report 2025**. Their finding: by 2030, **92 million existing roles will be displaced**, and 86% of employers expect AI and information-processing tech to transform their business within five years. About **39% of the skills you need to do your job today will change** by 2030.
Goldman Sachs put a bigger, scarier number in the air back in 2023: roughly **300 million full-time jobs worldwide are *exposed* to AI automation**, with two-thirds of U.S. occupations partially exposed and a quarter to half of the tasks in those exposed jobs automatable. Office and administrative support sits around 46% automatable; the legal field around 44%.
And it's not just projections anymore. Stanford's Digital Economy Lab — economist Erik Brynjolfsson and colleagues — looked at actual payroll data and found that workers aged **22 to 25 in the most AI-exposed jobs** (software developers, customer service, clerical work) have seen roughly a **13% relative decline in employment** since generative AI took off, with entry-level software roles down closer to 20%. The New York Fed clocked recent-graduate unemployment at 5.8%, and computer-science grads specifically at 6.1% — higher than the national average, in the field everyone was told was a sure thing.
That's the real story. Write it on the wall. Anyone who tells you nobody gets hurt is selling something.
- https://www.weforum.org/press/2025/01/future-of-jobs-report-2025-78-million-new-job-opportunities-by-2030-but-urgent-upskilling-needed-to-prepare-workforces/
- https://www.goldmansachs.com/insights/articles/how-will-ai-affect-the-us-labor-market
- https://digitaleconomy.stanford.edu/news/ai-and-labor-markets-what-we-know-and-dont-know/
---
## Now the part the headline cut off
Go back to that WEF report — the one with the 92 million displaced jobs. Read the next line, the one that doesn't fit on a thumbnail. Over the same period, the same forces are projected to **create 170 million new jobs.**
Subtract. That's a **net gain of 78 million jobs by 2030** — roughly a 7% net increase in employment, even as 22% of all jobs churn underneath. The destruction is real *and* it's the smaller number.
Goldman's "300 million exposed" works the same way. *Exposed* is not *eliminated* — it's the count of jobs where some tasks can be automated, and the same report projects that AI could add about **7% to global GDP, nearly $7 trillion a year**. Their base-case forecast for actual worker displacement during the transition is **6–7% of workers over roughly a decade** — a serious disruption, comparable to a bad recession spread over ten years, not the end of work.
This is the move the doom genre always makes: it reports the displacement number in 72-point font and buries the creation number, or skips it entirely. It's the exact same trick as the viral "30 minutes of Netflix equals four miles of driving" stat — a real-sounding figure, stripped of the context that makes it meaningful, traveling around the world before the truth gets its boots on.
- https://www.weforum.org/stories/2025/01/future-of-jobs-report-2025-jobs-of-the-future-and-the-skills-you-need-to-get-them/
- https://www.goldmansachs.com/insights/articles/generative-ai-could-raise-global-gdp-by-7-percent
---
## This is how technology has *always* worked. Always.
Here's the thing the panic conveniently forgets: we have run this exact experiment, over and over, for two centuries, and it has come out the same way every time. Not most times. Every time.
Start with the people who got it most famously wrong. In the 1810s, English textile workers — skilled weavers and spinners whose craft took years to master — watched the power loom and the spinning frame roll into the factories and understood, correctly, that their jobs were finished. So they did something about it: they grabbed hammers and smashed the machines. We named a whole word after them. **Luddites.**
They weren't stupid and they weren't lazy. They were scared, and they were *right* about the part they could see — their specific jobs really were destroyed. What they couldn't see was the part that hadn't happened yet. Mechanized textiles made cloth so cheap and so abundant that demand exploded, and the industry that grew up around it — mills, the mechanics who built and maintained the machines, cotton trade, shipping, garment-making, the entire modern clothing and fashion economy — ended up employing *far more people* than hand-weaving ever had. The machines they smashed created millions of jobs. They just couldn't see them from where they were standing. Nobody could.
That's the whole story, right there, in 1812. Now watch it happen again. And again. And again.
**Agriculture.** It destroyed the jobs of nearly half the country. In 1900, about **40% of the entire U.S. workforce worked on farms.** Today it's **under 2%.** Tractors, combines, and irrigation took those jobs and never handed them back. But the workforce didn't collapse — it exploded. The children of those farmhands became electricians, nurses, pilots, engineers, and a hundred things that never existed on a farm. It created far more jobs than it destroyed.
**The automobile.** It destroyed the entire economy built on horses — stable hands, blacksmiths, farriers, carriage and wagon builders, harness makers, the livery stables, all of it, gone. But it created more than it took, and it wasn't close: auto manufacturing, mechanics, gas stations, the parts and tire industries, road and highway construction, trucking and logistics, motels, diners, insurance, dealerships — and the entire suburban world we still live in today.
**The personal computer.** It destroyed the typing pool. It took the jobs of typists, stenographers, filing clerks, and the armies of people who moved paper for a living, and it made them obsolete. But it created more than it destroyed: the entire software industry, programmers, IT departments, systems administrators, hardware manufacturing, tech support — millions of jobs that had no equivalent before the machine showed up.
**The internet.** It destroyed travel agencies, video rental stores, the classified-ad business, the phone book, and a huge slice of brick-and-mortar retail. Made them obsolete almost overnight. But it created more than it took: web developers, e-commerce, digital marketing, cloud infrastructure, online businesses by the millions — and the entire creator economy that lets one person reach the whole planet from a laptop.
**The smartphone.** This one was brutal. It didn't just dent industries, it made whole product categories *vanish* — standalone GPS units, point-and-shoot cameras, MP3 players, PDAs, voice recorders, paper maps, payphones. Destroyed. But it created more than it destroyed by an enormous margin: the app economy, mobile developers, the app stores, mobile payments, the entire gig economy, and a generation of creators and businesses that live entirely inside the device in your pocket.
**Photography.** It destroyed the livelihood of the portrait painter and the sketch artist — the people who used to make a living capturing a likeness by hand. Made them obsolete. But it created more than it took: photographers, film and camera manufacturing, photo labs, photojournalism, and eventually the entire world of digital imaging. *And* here's the part worth holding onto — it made the old craft **more valuable, not less.** Once a perfect mechanical image was cheap and everywhere, the hand-painted portrait became rare, premium, prized. The thing made by a human, *because* it was made by a human, went *up* in value. Remember that one — it matters for what's coming.
And here's the data that ties it all together. MIT economist David Autor and his colleagues went through eight decades of U.S. Census records and found that **more than 60% of the jobs Americans worked in 2018 did not exist in 1940.** Sixty percent. App developer, UX designer, social media manager, solar installer, genetic counselor, cybersecurity analyst — none of it was on anyone's radar. The work that now feeds most of the country was *literally unimaginable* to the people whose jobs the machines were taking at the time.
**So here's the point. The whole point.** Every single time — the textile machine, the tractor, the car, the computer, the internet, the phone — the same thing happened. Jobs were destroyed. And more jobs were created than were destroyed. *Every. Single. Time.*
And every single time, the people living through the destruction couldn't see the creation coming, because it hadn't happened yet. The Luddites couldn't see it. The farmhands couldn't see it. Nobody ever sees it. The new jobs are always invisible from where you're standing — they *have* to be, they don't exist yet — which is exactly why every generation looks at the wave in front of it and concludes that this time, surely, there's nothing on the other side. And every generation has been wrong.
- https://www.strategy-business.com/blog/US-Farms-Still-Feed-the-World-But-Farm-Jobs-Dwindle
- https://www.nber.org/system/files/working_papers/w30074/w30074.pdf
---
## Now the honest part — because this time might actually be different
If I stopped there, I'd be doing exactly what I'm criticizing: cherry-picking the comforting half. So let's be straight, because an argument that can't survive its own counterevidence isn't worth making. There is a real reason to think AI doesn't slot neatly into the historical pattern, and it's not a vibe — it's structural.
**Every previous technology automated a *task* and left humans the rest of the job.** The tractor took the plowing and left the farmer everything else — and freed the next generation to go do something that didn't exist yet. The word processor took the typing but left the writer the writing. The spreadsheet took the arithmetic but left the analyst the judgment. Automation kept moving labor *around*; it rarely swallowed an entire job whole, because a job is a bundle of tasks and the machine could only ever grab some of them.
AI plus robotics threatens, for the first time, to grab the *whole bundle.* The thinking and the lifting — cognitive work and physical work, in one system. There's no obvious "rest of the job" to retreat into when the machine can, in principle, do all of it. That's a genuinely new thing, and the people waving it away are being just as lazy as the doomers.
Here's the part the optimists skip. The bet underneath every one of those success stories was simple: when a machine took your job, there was always *another field to run to.* The displaced weaver's grandkids went into industries that hadn't been invented yet. The displaced farmhand moved to the city and found work that never existed on the farm. The whole pattern depends on there being higher ground — some new kind of work the machine *can't* do, that humans climb toward while the machine takes the rung below.
AI plus robotics is the first technology that threatens to climb the ladder *with* us. When the machine can do the thinking *and* the lifting, the unsettling question isn't "what task does it take" — it's "what's the field we run to this time?" That's a genuinely new question, and anyone who answers it with a confident shrug hasn't understood it.
The evidence right now is honestly **contested**, and you should distrust anyone too certain in either direction. Stanford's "Canaries in the Coal Mine" study found that real entry-level decline. But other serious work points the other way: a 2025 NBER paper studying 25,000 workers across 7,000 workplaces found essentially *zero* effect on earnings or hours, and a Federal Reserve study of more than a million firms found no link between AI adoption and reduced job postings. Danish researchers tracking AI-adopting firms found only small effects. The truth is we are early, the data is noisy, and the people screaming "apocalypse" and the people scoffing "nothing-burger" are both ahead of the evidence.
So: AI's labor disruption is real, it may be structurally bigger than anything before it, and we do not yet know how big. All three of those things are true at once. Hold them.
- https://digitaleconomy.stanford.edu/news/ai-and-labor-markets-what-we-know-and-dont-know/
- https://stanfordreview.org/the-class-of-2026-is-struggling-to-find-jobs-and-its-not-because-of-ai/
---
## The real problem isn't jobs. It's who's left to buy anything.
Here's where the smart version of the fear lives — and it has almost nothing to do with unemployment statistics.
Suppose the pessimists are right. Suppose AI and robots really can do nearly everything, and corporations really do replace most of their workforce with machines that don't sleep, unionize, or take a salary. Run that forward one step further than the scary headline does, and you hit a wall that should terrify the corporations more than the workers:
**Who buys the products?**
A robot doesn't buy a car. A robot doesn't subscribe to a streaming service, book a vacation, or pick up takeout on the way home. The entire machine of modern capitalism runs on a loop: companies pay workers, workers spend wages, that spending becomes company revenue, repeat. Cut the workers out of the loop and you haven't built a hyper-efficient utopia — you've built a factory that produces goods for a market that can no longer afford them.
Henry Ford understood this a century ago when he doubled his workers' pay: he wasn't being generous, he wanted people who could actually buy a Model T. Automate away the paycheck and you automate away the customer.
This is the genuinely serious problem, and it's not a *jobs* problem — it's a *demand* problem. The threat AI poses to capitalism isn't that the robots take the work. It's that the robots break the loop that makes the whole thing run.
---
## How it could go sideways — one plausible chain
What follows is speculation. But it's the *reasoned* kind — a logical chain you can walk and argue with, not a prophecy. Here's one way the pressure could play out.
Demand craters because too many people have lost wages. Corporations — the ones who automated — suddenly can't sell what their flawless robot workforce produces. So they do what large industries always do when the floor falls out: they go to the government. *Save us. Subsidize us. We're too big to fail.*
But a government watching millions of angry, broke, formerly-employed citizens has a different math problem. Bailing out the companies that fired everyone is political suicide. The money, if it flows, flows to *people* — because the alternative is genuinely dangerous. History is brutally clear on what happens when large numbers of people are hungry, homeless, and watching their kids go without while a visible few own everything: it does not stay quiet. The French Revolution, the upheavals of 1848, the unrest of the 1930s — desperation at scale has toppled governments and remade nations before, and "it can't happen here" is the most expensive sentence in history.
Then comes the tax problem. Unemployed people don't pay much in taxes. So who's left holding the bill for a society where the corporations captured all the productivity and the public captured the unemployment? The corporations are. Where does a cash-strapped government with a furious population turn for revenue? To the only place the money went. How that confrontation resolves — whether through taxation, regulation, redistribution, or something uglier — is the open question of the century.
And there's a darker fork. If corporations end up controlling more real power than governments — more money, more infrastructure, more of the machines that everything now runs on — then the question stops being "capitalism or socialism" and becomes "who actually governs?" When governments fail or get hollowed out, history shows us what fills the vacuum, and it's rarely gentle. That's the dystopian branch. It's a real branch. I just don't think it's the likely one — and here's why.
---
## Why I don't think it ends in collapse
The doom scenario assumes the powerful can pull this off quietly. They can't — not anymore.
The thing that's genuinely different now isn't just the AI. It's that **information moves faster than power can suppress it.** Every coordinated screw-job leaves a trail. Whistleblowers exist. Journalists exist. Good people inside bad institutions exist, and they leak. You cannot run a hundred-year plan to immiserate the public when the public can document it in real time and route around the gatekeepers. The truth comes out — maybe late, maybe messy, but it comes out. That's a structural check the 1930s didn't have.
And there's a release valve the doomers ignore, because it doesn't fit the tragedy: we can just **pay people.** That's what Universal Basic Income is — a floor under the demand loop. And it's no longer a thought experiment.
Sam Altman's nonprofit OpenResearch ran the largest UBI study in U.S. history: **$1,000 a month, no strings, for three years**, to 1,000 low-income people, against a 2,000-person control group. The results demolished the laziest objection. Recipients **did not stop working** — they worked about 1.3 fewer hours a week and were *more* likely to job-search, more likely to start businesses, more likely to pursue education, and dramatically more likely to take a job they actually wanted instead of the first one that paid rent. They spent the money on food, housing, and transportation. The headline finding was *agency* — the freedom to make better choices.
It's not a silver bullet, and the lead researcher says so plainly: cash alone doesn't fix chronic illness, childcare deserts, or the cost of housing. But as a mechanism to keep the demand loop alive while the labor market re-sorts itself, UBI isn't a fantasy. It's a tested tool with three years of hard data behind it. If AI productivity really does add trillions to global output, the money to fund a floor is, definitionally, *there.* The fight will be over distribution, not existence.
- https://www.bloomberg.com/news/articles/2024-07-22/ubi-study-backed-by-openai-s-sam-altman-bolsters-support-for-basic-income
- https://www.newsweek.com/sam-altman-open-research-basic-income-openai-guaranteed-income-study-results-1928367
---
## The creator economy and the careers nobody can name yet
Here's the optimistic thesis, stated plainly: **the new jobs will come, the same way they always have, and a huge share of them will be people making things for other people.**
The creator economy is already proof of concept. Goldman Sachs values it at roughly **$250 billion today, doubling toward $480 billion by 2027.** Somewhere over **200 million people worldwide now identify as creators** — a job category that didn't meaningfully exist twenty years ago. That's the Autor "60% of jobs didn't exist in 1940" pattern happening *right now, in real time,* in front of us.
And — staying honest, because this series doesn't cherry-pick — the creator economy is not a free lunch. The earnings are brutally top-heavy: more than half of creators make under $15,000 a year, and only about 4% clear six figures. A world where "just become a creator" is the answer to mass displacement would be a world of a few winners and a long, thin tail of strugglers. Anyone selling the creator economy as a clean solution is doing the optimist's version of the doomer's lie.
But that's the wrong way to read it. The creator economy isn't *the* answer — it's a *preview* of the mechanism. When machines take over the production of physical goods and routine cognitive work, what's left, and what grows, is the stuff humans specifically want *from other humans*: craft, taste, story, presence, the handmade thing, the particular voice, the person who made it. Some of those creators will work entirely by hand. Some will use AI as a tool the way a photographer uses a camera or a musician uses a synth. Most will do both. The line between "human-made" and "AI-assisted" will blur into irrelevance, the same way "real photography" stopped meaning "film" — and the money will keep flowing to whoever makes something worth paying attention to.
We can't name most of these future careers yet, and that's not a weakness in the argument — it's the whole point. The farmer in 1900 couldn't have described "social media manager." The 1940 census had no word for "app developer." The jobs AI creates will be invisible from here until they're suddenly everywhere, and ten years from now we'll talk about them like they were always obvious. There will, in all likelihood, be *billions* of them, in forms we'd find as baffling as a telegraph operator would find a podcast.
- https://www.goldmansachs.com/insights/articles/the-creator-economy-could-approach-half-a-trillion-dollars-by-2027
- https://www.companieshistory.com/creator-economy-market-size/
---
## So, about that headline
None of this means the disruption is painless. It isn't. The farm-to-factory transition took generations and ground up a lot of real lives along the way — the long-run story was triumphant; the people living through the short run often weren't. The cost of these transitions has always fallen hardest on the people least able to absorb it, and the genuine moral failure isn't that technology changes — it's that we keep letting the displaced fall through the floor while we wait for the new jobs to show up.
*That's* the real conversation. Not "will there be work" — there will — but "who do we catch in the meantime, and how." That's an adult conversation about transition, training, and a demand floor. It is a completely different conversation from "AI is going to end work and end the world," and the panic deliberately blends the two so the fear can do the thinking for you.
So here's the whole sentence, finished:
Yes, AI will destroy jobs — millions of them, and some of them already gone. And every comparable technology in two hundred years of history destroyed jobs too, and every single time created more than it took, in forms nobody saw coming. AI might be different because it can take the whole job, cognitively and physically, and that's a real and unprecedented risk worth taking seriously. But the failure mode isn't robots doing the work — it's breaking the loop that lets people afford what the robots make, and we have tools for that, from UBI to redistribution to an economy that increasingly pays humans to be interestingly human. The truth travels too fast now to be buried, the productivity gains are real enough to fund a floor, and the new work — billions of jobs we can't yet imagine — is already starting to appear at the edges.
It won't be smooth. It never is. But "the end of work" has been predicted at every one of these turns, and the robots are, once again, near the bottom of the list of things to actually panic about.
Take a breath. Then go make something only you could make.
---
*Sources cited inline throughout. Displacement and creation projections from the World Economic Forum (Future of Jobs Report 2025), Goldman Sachs Research, and the Stanford Digital Economy Lab ("Canaries in the Coal Mine," Brynjolfsson, Chandar & Chen). Historical labor data from the U.S. Census Bureau, USDA, and Autor et al., "New Frontiers: The Origins and Content of New Work, 1940–2018" (Quarterly Journal of Economics, 2024). Contested current evidence drawn from Stanford, NBER, and Federal Reserve studies. UBI findings from OpenResearch's Unconditional Cash Study. Creator-economy figures from Goldman Sachs Research and aggregated market data. Numbers are rounded for readability; the evidence on AI's near-term labor impact is genuinely unsettled, and the political-economic scenarios in this piece are explicitly labeled as reasoned speculation, not forecasts. Every link is right there. Check them.*
Comments
Log in to comment