How China Is Building an AI-Era Labour Market Infrastructure

China’s latest AI-related employment policies point to a development that deserves more attention. China is not only encouraging firms and public agencies to adopt artificial intelligence. It is starting to build public infrastructure for monitoring how AI affects employment, identifying emerging skills needs, helping workers develop emerging capabilities, and using AI within public employment services.

The architecture is still emerging, and many of its most ambitious elements remain unproven. But taken together, China’s new employment strategy and a cross-ministerial “AI + Human Resources and Social Security” policy suggest a more systematic approach to governing workforce transition today.

Employment First: the wider policy framework

The starting point is the State Council’s Employment First Strategy for the 15th Five-Year Plan, issued in June 2026. Rather than treating employment as a downstream result of economic growth, the plan makes “high-quality full employment” an explicit economic and social policy objective.

It identifies mismatch between labour supply and demand as a central employment challenge and places “investing in people” alongside investment in physical assets. More importantly, it brings employment effects into broader economic decision-making: major policies, projects and productive-capacity decisions should be assessed for their impacts on labour demand, job creation and loss, and unemployment risk.

According to the policy, AI enters this framework in two different ways. The plan calls for new forms of human-AI collaboration and for policymakers to strengthen AI’s job-creating effects. But it also directs government to use AI itself to improve employment policy and public employment services.

The skills dimension is equally important. The plan calls for a national platform to improve matching between talent supply and demand, stronger feedback between graduate employment outcomes and education provision, and widespread adoption of a closed loop:

Job demand → skills training → skills assessment → employment services

The objective is to deliver more training and make it more responsive to actual labour-market demand and to connect it more closely to employment outcomes.

From employment services to AI employment-impact monitoring

A second policy makes the AI architecture much more explicit. In June 2026, four central agencies—including the Ministry of Human Resources and Social Security (MOHRSS) and the National Development and Reform Commission—issued the Implementation Opinions on Accelerating the Development of “AI + Human Resources and Social Security” Applications.

The document sets targets for roughly 20 AI-enabled application scenarios by 2026, exploration of around 50 high-value application pathways by 2027 and widespread sectoral AI adoption by 2030. Its full framework covers employment, social security, talent development, labour relations, HR services and public administration. From 2026, AI application scenarios must account for at least 40% of the annual Digital HRSS innovation-challenge list and must be built on the sector’s AI-model infrastructure.

But the most consequential provision is not a chatbot or automated government workflow.

The policy calls for an AI employment-impact assessment framework and a dedicated AI employment-impact survey mechanism. It also proposes AI-enabled monitoring of labour-market conditions and explicitly connects that intelligence to macro employment policy.

Traditional public employment systems often operate reactively: someone loses a job, enters the system and receives assistance. A functioning employment-impact monitoring system could potentially detect earlier signals, such as occupations undergoing rapid task restructuring, shrinking hiring demand, emerging skills shortages, or groups and regions facing concentrated transition risk.

The same policy proposes shortage-occupation maps, talent-demand mapping and analysis of regional, sectoral and occupational trends. In effect, China is also trying to build a form of skills intelligence: using data to understand which capabilities the economy needs, where shortages are emerging and where training resources should be directed.

What implementation is beginning to look like

The national framework is already appearing in local plans and funded projects, although implementation remains uneven.

Chongqing, a municipality in southwestern China, is developing a National AI Application Pilot-Scale Testing Base for human-resource services. Its AI plan describes vertical-model capabilities for job analysis and workforce matching, alongside intelligent recruitment, headhunting assistants and AI-enabled supervision of the human-resources sector. Its broader HR plan also proposes an “AI + human resources” incubation park and a pathway from research through pilot-scale testing to commercialisation.

Shanghai represents a different model. Its 2026–2030 Employment and Social Security Plan commits the city to participating in national HR-sector model development and training, expanding intelligent job matching, completing an “AI + mediation and arbitration” pilot, and building an intelligent labour-relations governance system. The plan therefore proposes AI not only for matching efficiency, but also for labour-rights administration and dispute resolution.

Longyan, a prefecture-level city in Fujian, offers a frontline public-employment-service example. Its “digital-intelligent employment” project has a planned budget of RMB 23 million, financed through a central public-employment-service capacity programme. Planned functions include AI job matching, targeted services for priority groups, personalised training and policy recommendations, and a citywide system for monitoring employment conditions and risks. Procurement had been completed by August 2026.

Together, these cases show three emerging implementation routes: AI infrastructure for the HR industry, AI-enabled metropolitan labour governance, and AI-enabled frontline employment services.

What problem is China trying to solve?

Three problems sit underneath these policies.

Structural mismatch. China’s employment strategy explicitly identifies the mismatch between workers and changing labour demand as a major challenge. AI is therefore being used not simply to automate employment offices, but to improve the information flowing between employers, workers, training providers and government.

Speed of technological change. Conventional labour statistics are often too slow or too aggregated to show how AI is changing tasks inside occupations. An employment-impact survey combined with more timely labour-market data could, in principle, provide policymakers with earlier signals.

Transition capacity. National AI strategies frequently concentrate on compute, models, data and industrial adoption. China’s emerging framework adds another layer: workforce transition infrastructure—the institutions needed to identify who is affected, understand which skills are required, help people retrain, and reconnect them with employment.

The unanswered questions

The ambition of the model should not be confused with demonstrated effectiveness.

Measurement is the first challenge. Separating AI-driven displacement from weaker economic demand, conventional automation or ordinary corporate restructuring is methodologically difficult. A company may never formally lay off workers but may reduce graduate hiring or redistribute tasks across existing roles—changes that conventional employment indicators may miss.

Data quality and interoperability are another constraint. The model depends on timely information moving across employment services, training systems, employers and government databases. China’s own 15th Five-Year HRSS digitalisation agenda calls for more real-time aggregation of employment, skills-training, labour-contract, dispute-resolution and labour-inspection data, suggesting that this infrastructure is still under construction.

There are also governance risks. Worker profiling, automated matching, social-security review and AI-assisted labour adjudication raise questions about bias, explainability, privacy and meaningful human oversight. The national policy explicitly calls for data protection, algorithmic security, ethical governance and AI risk monitoring, but the practical safeguards applied to high-stakes individual decisions will matter more than broad principles.

Finally, success should not be measured by how many AI scenarios are deployed. The stronger test is whether they improve actual transition outcomes: faster movement into viable work, better alignment between training and employer demand, broader access to employment support, and earlier identification of displacement risk.

What to watch

China’s approach is therefore best understood as an emerging attempt to connect labour-market intelligence, AI employment-impact monitoring, skills development, skills assessment and public employment services within a common administrative architecture.

The next question is implementation.

If employment-impact monitoring becomes credible, if the workforce loop produces measurable employment outcomes, and if local pilots generate interoperable standards rather than isolated systems, China could become an important case of how governments move from building AI infrastructure to building the workforce transition infrastructure that widespread AI adoption may require.