On September 4, The Economist ran a cover story in its finance-and-economics section:
The jobs apocalypse is postponed. An AI jobs boom is here.
The headline works as a pun — "postponed" applies both to mass layoffs that never materialized and to the dominant narrative getting flipped. The original is searchable on economist.com under "jobs apocalypse."
This piece is worth unpacking: what it's actually claiming, whether the data holds up, and why the Chinese-language tech community has barely touched it.
What The Economist Actually Said
The core argument in one sentence: the "AI will displace hundreds of millions of workers" narrative that dominated 2023–2025 has not materialized.
The Economist's September covers regularly fixate on the labor market — remote work in 2022, Gen Z's hiring struggles in 2023, mid-career unemployment in 2024. This year's installment follows the same playbook: using data to push back against prevailing anxiety.
The article draws on roughly three lines of evidence:
- Employment data from the U.S. BLS, Eurostat, and Japan's Statistics Bureau covering 2024–2026 shows no mass layoff wave directly attributable to AI.
- LinkedIn and Indeed job-posting data shows AI-related roles doubling year-over-year.
- Enterprise surveys indicate that after AI tool adoption, headcount budgets did not shrink — they grew, though the composition of roles is shifting.
On the surface, it's reassuring. But that last point — "the composition of roles is shifting" — is where the real story lives. More on that below.
Three Telling Details the Coverage Glosses Over
On Hacker News, this article was a strange presence: low engagement, few comments, never made the front page. HN is where the tech community is most reactive to macro narratives. If a contrarian take on "AI and the labor market" were genuinely explosive, engineers would pile in — either tearing apart the data or attacking the framing. The silence itself is a signal.
Three possible explanations:
- Engineers don't care about macro narratives — they'd rather see what Cursor shipped this week.
- The contrarian take isn't contrarian anymore — tech workers already know "AI replaces programmers" is overblown; they don't need The Economist to validate it.
- The data isn't convincing — no first-party numbers, no methodology.
I lean toward the third. The article's summary offers no concrete percentages, just "booming" and "growing." That kind of vague macro framing gets clocked immediately by a technical audience.
How the "AI Replaces Humans" Narrative Got Empirically Beaten
2023 was the year of anxiety. Three landmark predictions:
- Goldman Sachs projected roughly 300 million jobs globally would be affected by AI, with legal, administrative, and translation roles hit hardest.
- OpenAI CEO Sam Altman said in multiple interviews that AI would reshape customer service, translation, and entry-level coding (note: the specific "disappearing within 5 years" framing was a Chinese-media embellishment; his actual remarks were vaguer).
- Anthropic flagged white-collar risk in early policy papers but offered no specific replacement-rate figures.
Looking back from 2024–2026:
- BLS data shows the U.S. added about 2.5 million jobs in 2024, with continued growth in 2025 — well above pre-pandemic trends.
- The ILO's 2025 report puts global unemployment at roughly 5%, with no structural deterioration.
- Translation and entry-level customer service did contract, but the gap was filled by roles in e-commerce, cross-border livestreaming, and AI training data.
Why did the consensus miss so badly? My read:
Economists and AI companies conflated "technical feasibility" with "commercial deployment." A model that can translate doesn't mean a company will rip out its existing translation pipeline. A model that can write code doesn't mean a startup will trust it with core business logic. Deployment cost, compliance risk, reliability — those are the real moats.
The Goldman Sachs report was criticized at the time for measuring only "capability overlap" while ignoring organizational-change costs.
Don't Celebrate Yet: What Kind of Jobs Is AI Actually Hiring For?
The Economist says there's an "AI hiring boom." The composition of that boom is the actual story.
Pull up LinkedIn's 2024–2026 hiring trends and the top AI roles are:
- Data annotators — feeding RLHF pipelines, $15–25/hour, heavily outsourced to Kenya, the Philippines, and India.
- Prompt Engineers — exploded in 2023, cooled in 2024, now folded into product management as a sub-skill.
- RLHF annotators / AI trainers — heavy overlap with annotators; essentially "sparring partners" for models.
- AI data analysts — closer to traditional BI, just with a SQL + Python + LLM stack.
- AI product managers — real demand, but salaries aren't dramatically higher than general PM roles.
These roles share a few traits:
- Tightly coupled to frontier-lab release cycles — skill requirements shift every six months; career half-life is short.
- Bimodal pay — AI researchers at $1M+ total comp, annotators at $15/hour. Averages hide the structure.
- Displacement and creation are mismatched — translation, entry-level support, and junior copywriting are being cut; data annotation and AI training are being added. The former often requires a college degree; the latter doesn't.
So The Economist isn't wrong that "AI is hiring." But the overlap between displaced workers and new hires is close to zero. This isn't a labor-market victory. It's closer to a structural downgrade.
Editorial Take
Two judgments after reading the piece.
First, "contrarian" is becoming the new consensus. The 2023 mainstream was "AI replaces humans." The 2026 mainstream is "AI creates jobs." Both extremes are wrong — the reality is partial displacement, partial creation, structural mismatch. But contrarian narratives spread further than balanced ones. My guess for 2027: The Economist runs another cover, this time on the "AI hiring bubble bursting," because the sustainability of low-barrier annotation work is genuinely questionable.
Second, The Economist's blind spots matter especially for Chinese readers. Its data comes mainly from the U.S. BLS and Eurostat — almost nothing on China. And China's situation is arguably more complex: on one hand, domestic AI deployment is faster, so replacement rates in customer service, e-commerce operations, and junior programming may exceed Western levels; on the other hand, China hosts the world's largest data-annotation industry, meaning real AI hiring growth gets diluted by overall labor-market scale. Importing The Economist's optimism into a Chinese context is likely to be badly wrong.
One thing I'm unsure about: whether Goldman Sachs's 2023 prediction was simply wrong, or just early. Model capability keeps climbing, and organizational-change lag could run 5–10 years. Is the current "jobs boom" a steady state, or just a breather between two shocks? The Economist didn't answer that. Neither can I.