How China Is Responding to AI's Employment Impact

When artificial intelligence changes the way a business operates, the effects on jobs can be hard to see at first. A firm may hire fewer junior staff, shift some tasks to software or ask employees to work in new ways. None of these changes necessarily appears immediately as unemployment. For public employment agencies, this poses a practical problem: how can they spot a change early enough to help people affected by it?

China has begun to spell out its answer. On 23 October 2026, the State Council General Office published a policy opinion on adapting to AI development while fostering an employment-friendly economy. Issued on 16 October as Guobanfa [2026] No. 29, the document sets out 18 measures in six areas. It addresses the jobs AI may create as well as those it may disrupt, covering training, public employment services, labour protections, monitoring and the handling of employment risks.

The policy is recent, but not every part of China's response is new. Local governments have spent years building digital employment services and labour-market monitoring systems. Some of these are already in use; others are plans for the next stage. Understanding the difference gives international readers a more accurate view of how China's approach may work in practice.

A policy intended to create opportunities as well as manage disruption

The national opinion does not treat automation simply as a threat. It calls for employment growth in AI-related industries, supports smaller firms as they adopt AI, and encourages new occupations and collaboration between people and AI systems. Alongside those measures are proposals to manage the pace of deployment in labour-intensive activities, assess the employment risks of certain significant AI projects and help people whose work changes.

There is a second strand to this agenda. A June 2026 implementation opinion from the Ministry of Human Resources and Social Security (MOHRSS) and three partner departments describes how AI could improve human-resources and social-security administration itself. The programme includes public-service applications, data capabilities and staged goals through 2030. Those dates are targets, not evidence that every planned service is already in place.

Taken together, these policies address two sides of the same issue: helping businesses and workers adapt to AI, while improving the public services expected to support them.

Four ways the policy is meant to work

The October opinion covers more than monitoring. For clarity, its provisions can be read through four connected functions. This is a way of explaining the policy, not a formal four-stage procedure established by the government.

Assess risks as AI is introduced. Section 7 calls for employment-risk assessment in the review of certain AI application projects that could have a substantial effect on jobs. The aim is to consider the consequences for workers when deployment decisions are made, rather than only after redundancies occur.

Monitor changes in the labour market. Sections 16 and 17 propose sharing data across departments, developing regional and industry-level monitoring models, and conducting dedicated surveys. These would look at AI uptake, job creation and substitution, and changing skills requirements. The distinction between these measures matters: a job may be exposed to AI without its employer having adopted the technology, and adoption does not automatically mean displacement.

Prepare a response when risks emerge. Section 18 brings AI-related employment risks into arrangements for preventing and responding to large-scale unemployment. It also calls for planning and coordination when changes threaten labour relations or local employment stability.

Offer practical help to workers. Other provisions cover vocational training, career services, job placement, entrepreneurship support and protections for people in flexible forms of work. Monitoring only becomes useful to workers when the information leads to timely assistance.

These functions may run alongside one another, across several departments. For an employment authority, the test is whether an early warning reaches someone who can act on it.

Three local systems already at work

China's local employment systems do not all serve the same purpose. Three cases published through the official China Employment website in March 2026 illustrate how public authorities are collecting labour-market information, matching people with opportunities and delivering services. All three predate the October national policy. They should not be mistaken for evidence that the new policy has already produced an AI-specific monitoring system.

Quanzhou: turning labour-market data into alerts

In Quanzhou, Fujian province, the Quanjiuye platform offers a close look at how employment monitoring can function. Its smart-monitoring module began operating in April 2025. According to the official case account, it draws on information involving ten government departments and 22 categories of data, including social insurance, taxation, employment records and job-seekers' intentions. Warning terminals operate across 13 county-level jurisdictions.

The process includes automated screening followed by human checks. Alerts are then graded and passed to the relevant authorities. The account describes thresholds related to recruitment shortages and reductions in staff, as well as possible responses such as job-retention measures, training and labour coordination across regions.

This is an operating employment-warning system, with people reviewing the data before action is taken. Its published thresholds are local arrangements, not national AI-risk standards. Nor does the account show that Quanzhou can isolate job losses caused specifically by AI from those caused by other economic changes.

Ningbo: matching workers to vacancies and support

Yinzhou district in Ningbo, Zhejiang province, has taken a different approach. Its Employment Intelligence and Targeted Support service brings together information on job-seekers, vacancies, skills needs and available employment policies. The official account describes services across 22 subdistricts and townships, including vacancy matching, training referrals and more targeted notifications about support for which a person or employer may qualify.

The value of this example lies in what happens after a need is identified. Rather than leaving people to navigate separate systems, the service helps them find work, training or public assistance. It was operating before the October opinion and has not been presented as a dedicated service for AI-related redundancies.

Chengdu: using AI to make employment services easier to reach

In Jinjiang district, Chengdu, the Jinjiang Employment Supermarket brings online services together with local employment offices. The official case describes AI-assisted job recommendations, CV support, practice interviews, recruitment services and answers to policy questions, using tools that include DeepSeek-R1 and Doubao. Employment-service staff help with follow-up across 11 subdistricts.

This is a different use of AI in employment policy. The technology is not only something government may need to monitor; it can also help public agencies serve job-seekers. The case reports activity on the platform, but service volumes are not proof that AI itself created additional jobs.

Shanghai and Shenzhen: plans for the next stage

Local authorities are also setting out more explicit plans for monitoring the employment implications of new technology.

Shanghai's 15th Five-Year Plan for Employment and Social Security calls for closer monitoring of employment in key industries and enterprises, attention to the effects of AI and other technologies, and stronger early-warning and emergency-response arrangements. The plan also envisages wider use of AI in employment services. It is a formal policy direction, not a claim that every AI-specific monitoring capability is already operational.

Shenzhen's AI and Applications Action Plan (2026–2028) includes plans to strengthen the Pengcheng i-Employment platform, with functions covering employment analysis, labour coordination, unemployment-risk monitoring, job matching and skills services. Here too, a stated construction goal needs to be distinguished from a completed, evaluated system.

The five places therefore tell different parts of the story. Quanzhou demonstrates an existing warning mechanism; Yinzhou and Jinjiang show routes from information to employment assistance; Shanghai and Shenzhen describe further work they intend to undertake. They are not five branches of one unified national AI-employment monitoring platform.

What other countries can learn from this approach

China's emerging policy has a broad remit. It seeks to encourage new jobs, consider employment consequences when AI is introduced, watch for risks and provide help through public services. The local examples show why this requires more than good forecasts. Authorities need data they can use, staff who can act on it and services workers can actually reach.

For international policymakers, Quanzhou offers an example of how a warning can pass from data analysis to human review. Yinzhou illustrates how employment information can be linked to practical support. Jinjiang shows the role AI itself can play in delivering services. Shanghai and Shenzhen indicate where local policy is heading next.

The newer, AI-specific arrangements will take time to develop. That is worth stating plainly. What can already be seen is the combination of a national policy setting out responsibilities and local systems that provide some of the tools needed to carry them out. The lesson is not that every country should reproduce these platforms. It is that identifying a labour-market risk and helping a person through it are parts of the same public task.