Ai is eating the white-collar lunch: history says the pain is just starting
Silicon Valley keeps promising that artificial intelligence will “augment, not replace.” The ledger of history tells a nastier story. Between 1806 and 1820, hand-loom weavers—once the remote-work aristocracy of Britain—watched their real pay collapse by half while factory owners doubled output. They never recovered.
The first knowledge workers to fall
Weavers in 1800 looked a lot like junior analysts in 2020: flexible hours, above-average wages, social clout. Steam-powered looms did not politely nudge them up the skill ladder; they shoved them into the gutter. Census tallies are patchy, but wage books from Lancashire mills show the premium for hand-loom skills evaporated in fourteen brutal years. By 1820 a weaver earned 25 % less than the factory operative he once pitied.
Sound familiar? Goldman Sachs estimates 300 million full-time jobs world-wide could be automated “sooner than expected.” Notice the verb: automated, not augmented. The early casualties are not truck drivers or cashiers; they are paralegals, radiologists, entry-level coders—today’s equivalent of the starched-collar craftsman who thought machinery would stay downstairs.
What the novels saw that spreadsheets missed
Charlotte Brontë’s Shirley opens with frame-breakers smashing a shipment of looms under moonlight. Elizabeth Gaskell’s North and South locks a southern-bred pastor’s daughter inside a strike where the factory owner, John Thornton, insists he “cannot afford sentiment.” Dickens mints the era’s coldest line—"Are there no prisons?"—when asked why he won’t fund poor relief. None of these writers used the word “disruption”; they simply recorded the vertigo of watching status, income and identity vaporise overnight.
They also captured the lag. Productivity surged in the 1820s; wages did not catch up until the 1870s. A full two human generations ate the bitterness. If ai follows the same arc, a coder born in 1995 will be collecting social security before the median pay stub recovers.
Social capital buys only time, not immunity
Luddites could organise because they had savings, literacy, guild halls. Parliament still sent 14,000 troops to crush them. The same quasi-privilege now belongs to knowledge workers posting LinkedIn laments about GPT-4. The platforms listen, nod, and ship the next model anyway.
Here is the uncomfortable math: every previous wave of automation first attacked the middle layer—skills complex enough to be expensive, routine enough to be codified. ai is accelerating that playbook. The consultancy Accenture will happily charge a law firm $5 million to “integrate” an ai tool that, by next year, lets the same firm axe 30 % of its first-year associates.
The cotton gin lesson we keep ignoring
Eli Whitney’s 1794 invention was supposed to end slavery by making upland cotton unprofitable. Instead it exploded demand, entrenching bondage so deeply that America needed a civil war to dig out. The moral: technology does not care about the story we tell ourselves. It amplifies whatever incentive already pays the most. Right now, the highest returns lie in shaving payroll.
Federal Reserve data show white-collar job postings down 15 % since ChatGPT’s launch, even as factory openings rise. One summer internship at JP Morgan now draws 300,000 applicants for 400 slots. The queue feels Victorian.
History’s bargain: pain first, pie later
Yes, the steam engine eventually created more jobs than it destroyed. But the transition cost was paid in shortened lives, shuttered villages and children piecing quilts at midnight for a penny an hour. The pie arrived; the eaters were mostly the next generation.
Policy can shorten the lag—antitrust enforcement, portable benefits, robot taxes—or it can widen it. So far Congress is still asking if ai “hallucinates.” Meanwhile, Anthropic just raised $750 million to build models that write legal briefs faster than any first-year associate. Clock the gap.
Read the novels. They do not end with a town-hall where everyone agrees to reskill. They end with broken engagements, emigration, or early death. The lucky ones rewrite their stories in another city, under another name. The unlucky ones vanish into footnotes.
We are barely at chapter two of this cycle. The machinery is humming. The collars are still white—for now.