How China's AI race is shrinking job market, forcing workers into low-paid, manual roles

URL SCAN: How China's AI race is shrinking job market, forcing workers into low-paid, manual roles
FIRST LINE: For four years, Beijing-based Clare Zhang worked as a research analyst at a corporate advisory firm, helping companies assess technologies and competitors for projects they planned to list – until late last year, when she was told that artificial intelligence had reduced the number of analysts the company needed and could afford.

The Dissection

This is a field autopsy of the white-collar middle. The article documents the DT sequence: AI absorbs information retrieval, analysis, content production, coding, and customer-service work; firms reduce headcount or bring the work in-house; junior and mid-level roles disappear; experience loses scarcity value; displaced workers descend into lower-paid physical labor.

The decisive fact is not that AI assists workers. It is that teams are being compressed: work once requiring 10 people can be done by two or three. That is labor compression disguised as productivity. The article then places gauze over the wound with references to “transformation,” new AI jobs, a court ruling, and resilient manual sectors. Those are lags and niches, not reversal.

The Core Fallacy

The article confuses new tasks with replacement employment. The cited growth of roughly 90,000 AI application roles and 30,500 development roles shows gross creation, not whether displaced workers are absorbed, whether the jobs remain durable, or who captures the resulting surplus. AI can create specialist roles while destroying the broad career ladder beneath them.

It also mistakes “harder for AI to replace” for “economically secure.” Manual work, care, manufacturing, and maintenance are current physical bottlenecks. Under DT, they are lag defenses. They delay automation, but they do not restore the mass employment-to-wage circuit. If displaced cognitive workers flood those sectors, the result is a crowded low-wage reserve, not a new source of bargaining power.

The strongest contradiction is the article’s own evidence that employers demand immediate productivity and no longer support learning on the job. Reskilling is therefore not a general solution. AI fluency may preserve an individual temporarily as a servitor; it does not give that worker ownership or control of the AI capital generating the productivity gains.

The court ruling is another lag defense. It may restrict an explicit dismissal justified by AI, but it cannot force a firm to maintain unnecessary work. Employers can freeze hiring, cut pay, outsource, reorganize, or simply stop replacing departing staff. Law can obstruct the announcement of substitution without preventing substitution itself.

Hidden Assumptions

• AI-created jobs will scale sufficiently to absorb displaced workers.
• Occupation “transformation” will preserve headcount rather than conceal layoffs, wage cuts, and shrinking teams.
• Workers can move into AI-related roles despite employers eliminating apprenticeship pathways.
• Manual and care work will remain resilient long enough, and pay well enough, to function as a genuine alternative.
• Learning to operate AI preserves the value of prior expertise rather than making that expertise cheaper and interchangeable.
• Productivity gains will flow to labor through wages instead of accruing to firms, clients, and owners of AI infrastructure.
• Legal protections can preserve human employment in roles that no longer have an economic justification.

Social Function

Classification: partial truth, transition management, and ideological anesthetic, with a secondary element of elite self-exoneration.

It is partial truth because the cases and statistics clearly expose the first phase of cognitive automation. It is transition management because it reframes the collapse of the employment ladder as a sequence of individual pivots: use AI, retrain, move into care, maintenance, or manual work. It is ideological anesthetic because “new opportunities” and “transformation” soften the central question: how does the majority retain purchasing power after its labor is no longer necessary?

The focus on worker adaptability also shifts responsibility from capital ownership and distribution to individual skill acquisition. The article is not pure propaganda; its strongest evidence undermines its reassuring frame. But it reports a structural breakdown while narrating it as career turbulence.

The Verdict

China is not demonstrating that AI will create a better labor market. It is demonstrating the descent mechanism: cognitive work is de-bundled, middle layers are deleted, apprenticeships vanish, and remaining humans are sorted into a narrow AI-owner/controller class, a limited servitor class, and a crowded low-paid manual reserve.

Under the Discontinuity Thesis, this is direct evidence of P1 and early P3. It does not by itself prove that the entire post-WWII system has already died, but it shows the employment-to-wage circuit breaking at its center. Courts, retraining, and physical labor can delay the collapse. They cannot reverse it. This is an obsolescence report wrapped in the language of adaptation.