1.1.1 Job Displacement and Creation

In early 2025, a software engineer named Marcus logged into his work email to find a message that began with the phrase no one wants to read: "We're restructuring." His company, like dozens of others that quarter, was replacing his team—not with offshore contractors or cheaper labor, but with a suite of AI tools that could write code, debug errors, and deploy updates faster than he could finish his morning coffee. Marcus was one of 77,999 tech workers whose positions were eliminated across 342 company layoffs in 2025—a figure that captured the year through November, not merely the first half. He would not be the last.

We've been here before, of course. Or at least, we think we have.

Echoes of the Past

In 1764, a British carpenter named James Hargreaves built a wooden contraption that would change the world—and ruin a lot of lives in the process. His Spinning Jenny could spin eight threads at once, then eventually sixteen, then more. One person could now do the work of many. Within decades, hundreds of thousands of skilled weavers found themselves obsolete, their years of apprenticeship rendered worthless by a machine that hummed along without complaint, breaks, or wages.

The displaced workers didn't take it quietly. From 1811 to 1816, groups of textile workers calling themselves Luddites smashed the looms that had stolen their livelihoods. They were eventually suppressed, their protests crushed, and history remembers them mostly as a punchline—people who foolishly stood in the way of progress. But what we often forget is that those weavers were right to be afraid. Around 50,000 workers in England's textile sector lost their jobs, and those who found new work in the factories earned far less than the artisans they replaced. The transition wasn't smooth. It was brutal.

Fast forward to the 19th century, and steam-powered railroads were eliminating an estimated 200,000 jobs in horse-drawn transportation across Europe and North America. Then came the 1960s and 70s, when the automation of data processing eliminated approximately 400,000 clerical positions in the United States alone. Typists and bookkeepers—stable, respectable middle-class jobs—simply vanished.

Each time, society eventually absorbed the shock. New industries emerged. People retrained, adapted, moved. The economy grew. Looking back, we can see that in aggregate, more jobs were created than lost over those two hundred years. But that is cold comfort if you are the weaver standing in front of a power loom, or the bookkeeper being shown the door to make room for a computer.

The Numbers Game

So where do we stand now? The figures emerging from 2025 and early 2026 paint a picture that is both familiar and unsettling.

According to the World Economic Forum's 2025 Future of Jobs Report, 92 million jobs globally are projected to be displaced by AI and automation by 2030. In the United States alone, at least 23.2 million jobs have already seen 50% or more of their tasks automated—representing 15.1% of all U.S. employment. Meanwhile, Gallup's fall 2025 survey of more than 22,000 U.S. workers found that 12% use AI tools daily—nearly double the rate of two years prior—while Pew Research found that 19% use AI for at least some of their work, up from 14% the previous year. The following table captures the scale of projected displacement across key sectors:

Sector	Projected Impact
Technology	89,251 jobs cut in the first seven months of 2025—a 36% increase over 2024
Financial services	~200,000 jobs eliminated over the next three to five years
Manufacturing	Up to 2 million workers replaced by 2026; assembly line employment projected to fall from 2.1 million to 1 million by 2030
Human resources	44% of AI-using HR organizations applying AI to resume screening; overall AI adoption in HR reached 43% in 2026 (SHRM)

These figures are part of a broader pattern: 1,206,374 total job cuts in 2025—up 58% from 2024 and the highest annual total since the pandemic year of 2020. Nearly 55,000 of those tech-sector cuts were directly attributed to AI, according to consulting firm Challenger, Gray & Christmas. Even sectors long considered insulated are feeling the pressure. If you are filling out a job application right now, there is a good chance no human will ever read it.

Who Gets Hit Hardest?

Here is where things get particularly interesting—and troubling. It is not the jobs many expected.

For decades, conventional wisdom held that automation would come for manual labor first—factory workers, then truck drivers—while knowledge workers with college degrees remained insulated. That assumption has proven wrong. Research from Microsoft and other institutions shows that the occupations with the highest AI applicability scores, meaning those whose core tasks most closely align with what current AI systems can do, are not on the shop floor but in offices and cubicles. The contrast between the most vulnerable and most resistant roles is striking:

High AI Applicability (More Vulnerable)	Lower AI Applicability (More Resistant)
Interpreters and translators	Machine repair technicians
Writers and historians	Construction workers
Customer service representatives	Nurses and caregivers
Paralegals and junior accountants	Skilled tradespeople (plumbers, electricians)
Entry-level programmers	Physical and occupational therapists

Customer service roles—employing around five million people in the United States alone—are squarely in AI's crosshairs. So are paralegals, junior accountants, and entry-level programmers. Meanwhile, jobs requiring physical dexterity and on-the-spot human judgment are proving far more durable. The plumber who shows up to fix a burst pipe is, for now, considerably safer than the lawyer drafting the accompanying insurance claim.

There is another painful irony buried in the data. Young workers and those just starting their careers are being disproportionately affected. Entry-level positions—the traditional gateway to the middle class—are disappearing. Big Tech companies reduced new-graduate hiring by 25% in 2024 compared to 2023, and entry-level job postings have dropped 15% year over year. Nearly 50 million U.S. jobs disproportionately held by young workers are now considered at risk.

This matters beyond the immediate loss of income. Entry-level jobs are training grounds where workers learn workplace dynamics, professional communication, and how to manage competing demands. If an entire generation enters the labor force without those formative experiences, the long-term consequences for the talent pipeline—for who becomes the manager, the expert, the leader a decade from now—remain an open and concerning question.

But Wait—Aren't New Jobs Being Created?

Yes. And this is where the story becomes more complicated.

The same World Economic Forum projecting 92 million job losses by 2030 also estimates that 78 million new roles will emerge over that period. Some forecasts go higher still, projecting 170 million new roles this decade. AI is already spawning occupations that barely existed five years ago: machine learning engineers, AI ethicists, prompt engineers, synthetic data specialists, and AI policy analysts. In 2024 alone, around 119,900 AI-related positions were added—a figure that actually exceeds the roughly 12,700 confirmed AI-linked job losses recorded that year.

So is the ledger balanced? Not quite. The central problem is credential mismatch. Approximately 77% of those new AI jobs require a master's degree, and another 18% require a doctorate. For a displaced customer service representative with a high school diploma, or a junior accountant with an undergraduate degree, those new roles might as well exist on another continent. The skills gap isn't a gap—it's a chasm. Research on job postings shows that roles requiring new or emerging skills tend to pay around 3% more than comparable positions, rising to as much as 15% more in the UK and 8.5% more in the U.S. for roles demanding four or more new competencies. That premium is real, but capturing it requires access to education and retraining that is neither cheap nor quick.

There is also an important distinction that often gets lost in alarming headlines. When economists describe jobs as "at risk" from AI, they are rarely predicting that entire occupations will disappear overnight. They are identifying the share of tasks within a given role that AI can currently perform—and that distinction matters considerably. Radiology is the clearest illustration. For years, the profession was held up as a casualty-in-waiting of AI image recognition. What actually happened is that AI became highly effective at flagging potential anomalies in scans, which allowed radiologists to redirect their attention toward complex diagnoses, patient consultations, and treatment decisions that machines cannot meaningfully handle. The role evolved rather than evaporated. Similar patterns are emerging across legal research, financial analysis, and software development—AI absorbs the routine, repetitive components of a job while humans retain, and often spend more time on, the work requiring judgment, creativity, or interpersonal skill.

This helps explain a puzzling feature of the current moment: despite widespread anxiety about generative AI since its public debut in late 2022, the labor market has not experienced the catastrophic collapse in cognitive-job demand that many feared. Surveys of businesses show that while some firms have reduced headcount by deploying AI, a larger share report net increases in employment tied to AI adoption. The disruption is real, uneven, and accelerating—but it has not, so far, been a wholesale collapse.

The Shape of Things to Come

What does all of this mean? Are we heading toward sustained mass unemployment, or is this another wave of creative destruction that will ultimately improve living standards across the board?

The honest answer is that we do not yet know. What the historical record does tell us—along with what economists are finding in early AI-adoption data—is that this transition differs from previous industrial revolutions in at least one crucial respect: speed. Steam power took decades to spread through the economy. The personal computer revolution unfolded over roughly fifteen years. AI is moving faster. The lag between a new capability emerging in a research lab and companies deploying it at scale has compressed from years to months. McKinsey Global Institute research suggests that up to 30% of work activities could be automated by the end of this decade, a pace that outstrips any prior technological wave. That compression of time is the core challenge, because workers and institutions that once had a generation to adapt now face a far shorter window.

This speed asymmetry has distributional consequences. In earlier transitions, workers in vulnerable occupations often had enough warning—through union negotiations, early retirement packages, or visible changes on the factory floor—to begin planning an exit. Many of today's knowledge workers are receiving far less comparable notice. Moreover, the breadth of the current disruption is unusual: prior waves of automation primarily targeted a relatively narrow band of physical or routine tasks. AI, because it operates on language, pattern recognition, and reasoning, touches a far wider swath of the economy simultaneously, from entry-level legal work to financial analysis to creative production.

What this means in practice is that the adjustment burden falls disproportionately on individuals rather than on institutions. Retraining programs, where they exist, are often short-term and disconnected from employer needs. Community colleges and workforce development agencies are under-resourced relative to the scale of change underway. Researchers and institutions—from the IMF to the OECD—are increasingly calling for policy frameworks that can accelerate and broaden access to the new skills AI is creating demand for, not only at the graduate level, but across all education and experience tiers.

History offers a useful, if imperfect, frame of reference. Previous technological revolutions ultimately created more total jobs and more total wealth than they destroyed, but the transition periods involved genuine suffering, dislocation, and social upheaval that persisted for decades. The question for AI is not whether net job creation will eventually occur—the balance of evidence suggests it will—but whether the institutions we have built are capable of distributing those gains broadly and quickly enough to prevent the transitional pain from becoming a defining feature of the era.

Key Takeaways

Historical pattern, accelerated pace. Technological displacement of workers is not new—the Spinning Jenny, steam railroads, and computerized data processing all eliminated large categories of employment before creating new ones. What distinguishes AI is the speed of adoption, which leaves workers and institutions far less time to adapt than previous transitions allowed.
Scale of current disruption. By early 2026, 23.2 million U.S. jobs have seen at least half their tasks automated. Globally, the World Economic Forum's 2025 Future of Jobs Report projects 92 million jobs displaced and 78 million new roles created by 2030—but the gains and losses are not landing on the same workers or communities.
Knowledge workers, not just manual workers, are vulnerable. Contrary to long-held assumptions, AI most directly targets tasks in white-collar roles—translation, writing, customer service, legal research, accounting—while physically demanding or judgment-intensive trades prove more resistant.
Entry-level jobs are the most acute near-term casualty. The disappearance of entry-level positions threatens not just immediate income but the experiential ladder that workers use to build professional competence over time, with uncertain consequences for the next generation of skilled workers and leaders.
Job transformation, not just elimination. Many roles are evolving rather than vanishing outright, as AI absorbs routine tasks and humans concentrate on higher-judgment work. The radiologist, the lawyer, and the programmer are already experiencing this shift.
The new jobs exist, but access is unequal. Roughly 95% of newly created AI roles require a master's or doctoral degree, making the transition far harder for displaced workers without advanced credentials. The skills premium is real but unevenly accessible.
Speed and breadth define the central challenge. The core risk is not permanent technological unemployment but a transition period whose pace outstrips the adaptive capacity of individuals, educational systems, and social safety nets—requiring deliberate policy responses rather than patience alone.

Sources:

Last updated: 2026-03-07