Every generation has had its version of this fear. When the power loom arrived in 18th-century Britain, textile workers smashed the machines in the streets. When the automobile replaced the horse, people worried about every stable hand, blacksmith, and harness maker in the country. When computers arrived in offices in the 1980s, economists predicted mass unemployment among secretaries and clerks.
In every case, the fear was understandable. In every case, the long-run outcome was more jobs than before. And in every case, the transition period was brutal for the people caught in it.
AI is different in important ways. But it is not as different as either the doomsayers or the optimists insist. The evidence — and there is now quite a lot of it — tells a story more nuanced, more uneven, and ultimately more hopeful than the headlines suggest. Here is what we actually know.
The Numbers Everyone Is Citing — and What They Actually Mean
The World Economic Forum's 2025 Future of Jobs Report surveyed more than 1,000 employers representing over 14 million workers worldwide. Its headline finding deserves to be stated precisely — and in full.
What Is Actually Happening Right Now
The empirical data from 2025 and early 2026 is now substantial enough to move beyond projection and into measurement. The picture it reveals is neither catastrophe nor smooth sailing.
In the technology sector, the first six months of 2025 saw nearly 78,000 tech jobs attributed to AI-related layoffs — concentrated in entry-level and routine positions: QA testers, technical writers, Tier 1 support. Simultaneously, AI engineering, platform, and AI product roles saw salary growth of 20–40% as demand outpaced supply. For many workers, AI is currently augmenting rather than replacing output.
| Role / sector | Risk level | What is actually happening |
|---|---|---|
| Data entry & admin | Critical | 7.5M jobs projected eliminated by 2027. Already well underway. |
| Customer service | Critical | 80% automation reached by 2025. AI chatbots saving businesses $8B annually. |
| Junior legal / paralegal | High | AI reviews 500-page contracts in minutes. Firms cutting junior intake. |
| Financial analysis | High | A third of transaction-handling roles in finance already automated. |
| Software development | Mixed | Junior roles slowing; AI-focused & senior roles up 20–40%. Skill split widening. |
| Healthcare / clinical | Mixed | AI matches radiologists on imaging. Admin at risk; clinical & patient-facing roles growing. |
| Teaching & education | Low | AI augmenting, not replacing. New AI curriculum-design roles emerging. |
| Skilled trades | Low | Electricians, plumbers, carpenters — physical dexterity strongly resists automation. |
The Part Nobody Talks About: The Skills Premium
Buried beneath the displacement headlines is a finding that should command far more attention: the wage premium for AI skills nearly doubled in a single year.
"Technology destroys jobs, but not work." — US National Commission on Technology, Automation, and Economic Progress, 1966.
According to PwC's 2026 Global AI Jobs Barometer, the skills needed for the most AI-exposed jobs are changing more than twice as fast as those for the least exposed roles. AI-exposed junior roles are now 7× more likely to demand senior-level skills, compressing the time to advancement for workers who adapt. The professionals thriving are not those competing with AI. They are those using AI to do things previously impossible at their level — and developing the judgment AI cannot replicate.
What History Actually Promises — and What It Doesn't
The standard reassurance during every wave of automation anxiety is: technology has always created more jobs than it destroyed. This is largely true at the macro level. And it is almost entirely useless at the individual level.
| Technology wave | Short-run | Long-run | Who paid the transition cost |
|---|---|---|---|
| Industrial Revolution | Craft jobs destroyed; real wages fell | Massive productivity growth; new industries | A generation of skilled craftsmen |
| Agricultural mechanisation | US farm labour fell 20M → 4.5M | Food abundance; urbanisation; services | Rural communities |
| Computing & offshoring | Manufacturing & clerical hollowed out | Internet & software economy | Middle-skill workers |
| AI (2022–present) | Entry-level cognitive roles disappearing | Projected +78M net jobs | Routine cognitive workers & new graduates |
Every technology wave created more in the long run than it destroyed — but the transition cost was always paid by specific people, not spread evenly. Electricity transformed manufacturing, but the full productivity gains took nearly 40 years, because factories had to be redesigned around it rather than swapping steam engines for electric motors one-for-one.
The Honest Answer
The macro answer is cautiously optimistic. More jobs will be created than destroyed, and the AI boom is itself generating enormous employment — data centres, hardware, AI development, governance, and the industries that infrastructure makes possible.
The micro answer is more sobering. The transition will not be smooth or equitable. The workers most at risk are those in routine cognitive roles — the entry-level positions that have historically been the first rung on the career ladder.
The individual answer is the most actionable. The professionals thriving are not the ones competing with AI. They are the ones using AI to do things that were previously impossible at their level — and building the empathy, judgment, creativity, and communication that AI cannot replicate.
Opportunities: a 56% wage premium for workers who acquire AI skills — the strongest predictor of earnings growth through 2030. 650,000 data-centre jobs projected by 2026, 340,000 currently unfilled. AI-exposed junior roles are 7× more likely to demand senior-level skills.
History did not give the industrial era to the workers who wove cloth fastest by hand. It gave it to those who understood what the machines could and could not do — and built something new around that understanding.
The lamplighters who smashed the electric lamps were not wrong to fear the change. They were wrong about what to do next.
Next in this series: The State of Model Training — what it costs, who's winning, and why it is reshaping geopolitics, energy, and the future of AI power.