How AI Shakes Up the Labor Market
Source
- Article: Cai Fang (蔡昉), Proactively Addressing the Employment Impact of Artificial Intelligence.
URL = https://www.qstheory.cn/20260619/4abab8e31a95432aa919864d426fc10d/c.html
Commentary by the Taejae Future Consensus Institute, World Research Trend vol53_260814.
Discussion
Ruoying Yu:
"An AI Employment Warning in the Party’s Leading Theory Journal In June, the Chinese Communist Party’s flagship theoretical journal Qiushi (求是) published a notable article by Cai Fang, one of China’s top labor economists and a member of the Chinese Academy of Social Sciences. In this piece, Cai discussed the employment impact of artificial intelligence as a pivotal factor in China’s public welfare and social policy agenda.
Cai argues that the employment repercussions of AI could exceed those of any previous technological shock over the past few centuries, including the Industrial Revolution. More strikingly, he warns that the disruption may “not even have truly begun yet.” The effects of AI could quietly build up unnoticed before suddenly becoming visible on a much larger scale.
In Cai’s view, the core risk lies in the mismatch between the speed of technological change and the speed of institutional adaptation. AI is advancing rapidly, while employment, education, and social security systems are largely designed around relatively stable jobs and gradual technological change. If institutions fail to keep pace as AI shakes up the value of jobs and skills, the consequences may extend far beyond unemployment and impact education, career development, income and welfare. Institutional lag is hardly a new problem, but Cai’s argument is that AI fundamentally alters the scale and speed of the mismatch.
In this case, the forum matters as well. Qiushi is not an academic journal, but one of the CCP’s most authoritative outlets for political and policy discourse. In this sense, Cai’s proposals should not be read as an official policy blueprint that has already been endorsed by China’s top leadership.
However, the publication of this piece in Qiushi suggests that AI-related employment disruption has shifted from the margins of technology policy to become a core consideration in China’s social and public welfare agenda.
The End of Entry-Level Jobs and the Breakdown of Early Career Pathways
Cai describes the AI shock as an exacerbation of the “structural employment contradictions” China is already grappling with. As large language models approach average human performance in a growing number of white-collar tasks, the most vulnerable groups are likely to be the first victims of displacement.
In other words, young people who are just entering the labor market will bear the brunt of the AI-induced shock.
The central problem is not just job loss, but disruption in skill formation. Entry-level workers have traditionally learned how organizations function and accumulated professional knowledge by performing foundational tasks such as research, drafting, and information processing. If AI increasingly takes over these tasks, young workers may lose not only first job opportunities, but also the pathways through which they can develop into mid-level and senior professionals.
As the value of entry-level work declines, the link between education and employment may also weaken. The traditional “sheepskin effect,” whereby a university degree serves as a signal of competence, becomes less reliable, and employers may find it harder to infer actual skills from educational credentials alone. Frequent job changes, short-term contracts, and irregular employment could therefore become increasingly common features of labor market entry for younger workers.
These changes also interact with the rise of platform work and flexible employment. As jobs are created and destroyed more rapidly, workers may find themselves in a state of perpetual transition rather than temporarily between jobs. Labor institutions are largely built around formal contracts, social insurance, fixed working hours, and stable workplaces, which makes them poorly equipped to protect workers engaged in fragmented employment across a number of platforms. And as job matching and task allocation in turn become increasingly dependent on algorithms, the bargaining power of workers could be further eroded.
Older workers grappling with the widening digital divide are likely to be another vulnerable group, while robotics and embodied intelligence may increasingly affect blue-collar jobs in manufacturing and services. The broader implication of Cai’s analysis is not that AI will eliminate certain occupations, but that it could destabilize the entire “employment life cycle” from labor market entry and skill formation to income and social protection.
Changing the Direction of Technology and the Choices Firms Make
The first part of Cai’s proposal is to steer AI development toward what he calls an “employment-friendly” trajectory. A key concept in his discussion is duibiao (對標), which can be understood as aligning technological development with clearly defined social benchmarks or policy objectives. While conventional debates about AI alignment typically focus on ensuring AI systems follow human values, Cai is primarily concerned with changing the incentives of those who develop and deploy AI.
Regulation, subsidies, public procurement, and research funding should reward technologies that complement human labor and create new jobs instead of those that simply replace workers.
The rationale behind this is that the interests of individual firms do not necessarily coincide with those of society as a whole. For a single company, reducing headcount and lowering costs through AI may be perfectly rational. But if every firm makes the same choice, employment and labor income would be decimated, with ensuing effects on household consumption. What makes sense at an enterprise level has the potential to produce a socially damaging “fallacy of composition.”
This does not mean the state should directly pick individual technologies. Rather, Cai argues that policymakers should change the incentive structure that shapes which technologies firms find profitable to develop and adopt. Technological performance should not be judged solely on cost savings or productivity gains, but also on whether it creates jobs and expands the productive capacity of workers.
The second part of Cai’s response is to rebuild the pathways through which young people accumulate professional expertise. If AI erodes traditional entry-level work, young workers can noblonger rely on the old model of entering a profession and gradually learning on the job. Education, training, and skills certification therefore need to become part of a more integrated system.
In Cai’s proposal, human capital investment is not limited to vocational training immediately before labor market entry. He places particular emphasis on the “first 1,000 days” from pregnancy through roughly the first two years of a child’s life. The cognitive capacity, emotional stability and social interaction skills developed during this period, he argues, are difficult to fully compensate for through later education. Accordingly, Cai calls for greater public investment in maternal and child health, childcare, and preschool education along with more universal access to education up to the upper- secondary level. In the AI era, addressing inequality means reducing disparities not only through adult retraining, but also from the very beginning of life.2
Micro-Credentials as a New Form of Skills Recognition
The content of education must also change. In a world where AI can process enormous volumes of knowledge, years of schooling and memorized information can no longer serve as sufficient measures of individual competitiveness. Instead, our ability to understand and use new technologies, solve problems, communicate and empathize, collaborate with others, and exercise judgment in complex situations will be of greater value.
Under this paradigm, the role of universities is to cultivate the underlying capacity to keep learning, absorb new knowledge, and adapt to a changing work environment. Serving as short-term vocational training centers will not suffice.
Lifelong learning and on-the-job training after the completion of formal schooling need to become institutionalized. Since a university degree can no longer guarantee competence across an entire career, both the state and employers should assume greater responsibility for ensuring that workers can return to education and training whenever necessary. “Micro-credentials” are one of the mechanisms Cai suggests for making newly acquired skills visible in the labor market. Specific competencies gained through shorter courses and training programs can be certified step by step and built up over time, allowing workers to make use of them when changing occupations or careers. Taken together, Cai’s proposals point toward a life-cycle model of human-capital development that stretches from early childhood investment to formal schooling, workplace training, and micro-credentialing.
A Gradual Move Toward Decoupling Social Protection & Employment
Cai’s third suggestion is to prevent individuals from bearing the full cost of technological disruption and occupational transition. He proposes four forms of “decoupling,” namely the gradual separation of social protection from employment status, labor compensation from individual productivity, basic public services from place of residence, and social security benefits from individual contribution histories.
The underlying logic is that it is becoming increasingly difficult to distinguish individual responsibility from the effects of technological change. In the AI era, there is no way for workers to easily determine which occupations will survive and which will be replaced. Nor is it always possible to discern whether a person has been pushed out of the labor market due to insufficient effort or because technological change has reduced the value of their skills. The risks associated with unemployment, retraining, and career transition should therefore not be borne entirely by individuals. A guaranteed minimum level of income and access to public services should give people the capacity to retrain and try again.
Cai identifies universal basic income, a living wage, and non-contributory social pensions as possible reference points for the future evolution of China’s social security system, but stops short of advocating the immediate introduction of a basic income for the entire population. In consideration of China’s current fiscal constraints, he takes a more gradual approach. Basic income can serve as a long-term benchmark while the government first bolsters existing welfare systems and expands the scope of social protection.
These three proposals ultimately connect three distinct questions.
- What kinds of AI should society
encourage firms to develop?
- How should people prepare to adapt to the changing labor market?
- And how should society guarantee another chance to those whose jobs are disrupted by technological
change?
Cai’s broader point is that the AI-induced employment shock cannot be addressed through technology policy alone. It requires simultaneous changes in education, social protection, fiscal policy, and the way in which states operate."