Will AI take jobs or create jobs? Experts and corporate executives assess the employment landscape

The policy side has already stepped in. The State Council has made systematic arrangements for building an employment-friendly development model that adapts to AI technology, explicitly aiming to amplify AI's positive effects in creating jobs, empowering positions, and improving quality, mitigate substitution impacts, and achieve inclusive development in which technological progress and employment promotion are coordinated. The 15th Five-Year Plan also proposes accelerating the construction of a professional, digital, law-based, and internationalized human resources service system.

After the national top-level design was introduced, market frontlines are also paying close attention to how the new generation of AI, centered on generative large models and agents, will reshape the labor market. Recently, multiple experts, scholars, and corporate executives conducted analysis around this topic, outlining from theoretical frameworks to industry data, and from policy support to corporate practice, the real picture of the employment landscape under the impact of AI, and exploring countermeasures.

Substitution or creation? The employment landscape is being reshaped

Changes in the AI field have triggered innovations in production methods, work methods, service methods, and consumption methods, and have also brought new changes to employment and the human resources market.

Yang Zhiming, former Vice Minister of the Ministry of Human Resources and Social Security and a specially invited researcher at the Counselors' Office of the State Council, used a set of data at the recently held 2026 HRA Human Capital Sustainable Development Ecology Conference to outline four major trend changes in current employment:

Skilled talent is shifting from a quantity-oriented model to a quality-oriented model. Last year, China had 220 million skilled workers, but only 74 million highly skilled workers, accounting for 33%. In Germany and Japan, manufacturing powerhouses, this proportion is 50%. Accelerating the development of high-level talent on a large scale and at multiple levels is a strategic measure for the country's economic transformation.

Newly added skilled talent is shifting from unlimited supply to limited supply. The peak appeared in 2010, with 12.45 million new additions, and it has declined year by year since then. Last year it grew by only 0.4%, and in the first three quarters of this year it was in the 0.3%–0.5% range. Skilled talent is becoming a precious human resource, no longer an inexhaustible reservoir.

The new generation and the internet-native generation have become the main workforce in employment. Those born in the 1980s and 1990s are the new generation, while those born in the 2000s and 2010s are the internet-native generation. Relying on express delivery, food delivery, ride-hailing services, platform services, technical services, and household services, they reached 85 million last year and are expected to reach 90 million during the 15th Five-Year Plan period, becoming the largest pillar industry in China's occupational field.

Every technological revolution reshapes the employment landscape. The first industrial revolution replaced manual work with machines, the second replaced individual laborers with assembly lines, and the third replaced part of human labor with automation. Today, AI technologies represented by large models, advanced algorithms, agents, and embodied intelligence are changing production and life in a more thorough way.

The substitution effect means that when AI works at lower cost and higher efficiency, companies are motivated to replace people with AI, which at the macro level may reduce the scale of employment. The restoration effect has three layers: first, the productivity effect—AI improves efficiency, and companies instead need more workers; second, the scale effect—AI lowers costs and prices, and demand expansion drives job creation; third, the new occupation effect—AI spawns new jobs, and the employment scale of workers with complementary skills expands.

The article proposes that the general consensus of most theoretical research is that AI's impact on employment has significant long-term dynamic effects, basically presenting a judgment of long-term optimism and short-term pessimism. At present, two types of employment risks need to be guarded against: first, skills upgrading cannot keep up, causing structural mismatch in job demand; second, the unemployment risk that may be triggered because AI's creative effect has not yet been fully released.

The two scholars suggest proactively establishing a skills demand forecasting mechanism, moving education supply and skills training forward, and enabling the workforce's skills structure to keep up with technological iteration. At the same time, improve the employment safety-net guarantee, promptly match affected unemployed groups with assistance plans and employment guidance, raise the "person—job" matching rate, and use subsidies and job-stabilization policies to encourage companies to lay off fewer workers.

The General Office of the State Council recently issued the Opinions on Adapting to the Development of Artificial Intelligence Technology and Helping Build an Employment-Friendly Development Model (hereinafter referred to as the Opinions), deploying 18 major tasks in 6 areas, including strengthening AI's employment creation effect, building an employment-friendly AI upgrading path, and improving workers' digital literacy and employability.

At present, AI technology is already releasing employment demand by creating entirely new industry tracks. It not only directly spawns emerging occupations such as AI technicians, trainers, algorithm testers, and generative AI system application personnel, but also, by empowering the digital transformation of traditional industries, drives new forms of employment such as ride-hailing drivers, food delivery riders, and e-commerce livestreamers.

The Opinions propose improving workers' digital literacy and employability, collaboratively cultivating AI innovative talent, and implementing a special AI training plan to encourage college graduates and other young people, migrant workers, unemployed persons, and others to participate in AI-sector employment and entrepreneurship training programs.

Many companies have now launched AI platforms. How can employees feel it is empowerment rather than being monitored and replaced? Many companies are falling into a dilemma: under Token cost pressure, including AI use in KPI assessments risks unaffordable costs and easily triggers abuse and resistance; not including it raises concerns that employees will not use it.

Wu Wenqi, Vice President of Human Resources for Cisco Greater China, said Cisco built an AI assistant platform five years ago, with all AI tools running on the platform to solve data governance issues. Employees can choose the world's most advanced large models for comparison, and the most important thing is to make employees feel it is easy to use and serves their own needs.

“Cisco has never put AI usage into employee KPIs; it only regularly looks at usage frequency, does not track how many times each individual uses it, and only periodically shares who uses it well and how they use it well. Under Token cost pressure, the system prompts the remaining quota when questions are complex, guiding employees to ask more precise questions rather than relying on assessment pressure, thereby avoiding Token abuse,” Wu Wenqi said.

Liang Xue, Operations and HR Head of Yanming Biotech, said that AI still has very limited use in innovative R&D for entirely new targets and entirely new mechanisms; AI is more about improving efficiency, molecular screening, and offering some help in exploratory directional work, but it cannot replace these core scientific judgments. In actual work, the company has also made clear what can be asked of AI and what absolutely cannot, set red lines for core data, and at the same time encourages employees to think through the question clearly before turning to AI.

Liang Xue believes that AI currently cannot provide standard answers to non-closed-loop problems; AI cannot create from nothing, but people can achieve innovation from 0 to 1, so the company has established a fault-tolerance mechanism, hoping everyone will persist in or retain their own thinking and creativity.