[Big read] China’s young workers pay the price of AI before reaping the gains

Two months ago, Anthropic founder Dario Amodei, whose start-up is at the forefront of artificial general intelligence (AGI), effectively started a countdown on the “lifespan” of certain human jobs: programming roles could be completely replaced by artificial intelligence (AI) within six to 12 months, while half of entry-level white-collar positions may disappear within one to five years.

Across the ocean in China, fast-iterating AI systems are turning the once sci-fi idea of silicon employees replacing carbon-based humans into something uncomfortably close to reality.

Shandong-based third-party programmer Qin has been in the industry for over a decade and is among the first to feel these changes. He told Lianhe Zaobao that a project quoted at 10,000 RMB (US$1,450) in 2021 now only fetches around 5,000 RMB.

Even so, clients still challenge him: “AI can write it for free — why are you charging so much?”

AI has not only halved prices, but has also rapidly eroded front-end development work, such as webpages and mini-programmes that Qin used to receive. Clients now use AI to generate a draft, then hand the code to him for revisions. As prices are driven down, the nature of his work has shifted from writing to fixing code. “It feels like I’m working for AI,” he said.

Even programmers at large tech firms, whose jobs are relatively more stable, increasingly feel a looming sense of insecurity. Shi Wenyu (pseudonym), a senior programmer at a major internet company in Beijing, said that over the past year, large AI models have made significant advances in natural language understanding in software development and can now assist in almost every stage of the process. Over 30% of the code in his company is generated by AI.

However, the underlying architecture of large-scale applications in major firms is more complex, and AI is still unable to fully comprehend it for now. Nevertheless, Shi believes that the focus of programmers’ work has already shifted, and a key task for the future may be to make code architectures more AI-friendly.

To some extent, being more AI-friendly also implies that more work will be replaced by AI. Shi remarked somewhat helplessly, “There’s no point in being anxious, because this trend is irreversible.” He estimates that around 90% of programmers may be eliminated within the next three years.

In a report released by Anthropic in early March on the impact of AI on the labour market, besides the programming fields of computer and mathematics, business and finance, office administration, management, law, and arts and media were also identified as areas where AI could handle up to 80% of tasks.

In China, the impact of AI on employment has spread from technical roles to a broader range of sectors. Entry-level white-collar workers who primarily rely on computers as their production tools are experiencing a rapid restructuring of work patterns and career pathways.

Zhu, 25, who graduated two years ago with a degree in Chinese language and literature, now works as an administrative assistant at an engineering firm in Guangdong. Currently, 50-60% of his work is assisted by AI tools, while almost all the remaining tasks that AI cannot handle involve interpersonal interactions.

Over the past two years, after trying most of the popular Chinese and overseas large models, Zhu has settled on three that he finds most effective, and he readily explains the subtle differences between them during our interview.

However, with his monthly salary of 7,000 RMB, Zhu is keenly aware that “the credit cannot all go to AI”, and only tells his supervisor that he uses AI to correct typos and punctuation. “If AI does too much, my boss will think I’m underworked and not worth the salary,” he revealed.

Li Li, 24, who has been working for just over a year in new media operations in Anhui, has already had a pay cut because AI has taken over part of her workload. She explained that the main responsibilities of her role — copywriting, video editing and trend curation — can now all be completed or optimised by AI.

After AI reduced her workload, Li’s office attendance dropped from five days a week to three, and her salary fell from over 5,000 RMB to just over 3,000 RMB. To reduce living costs, she moved to the outskirts of Hefei, paying a monthly rent of 800 RMB. Feeling an increased risk of unemployment, she started submitting job applications again last month.

For job seekers who have yet to enter the workforce, the challenges are even more severe. Lu, 20, is one of the 12.7 million Chinese university graduates set to leave campus this summer. As AI replaces entry-level roles, the number of available positions for her has declined. Meanwhile, with AI augmenting existing jobs, the skill requirements for those positions are steadily rising.

Although Lu is a humanities graduate, many of the positions she applied for require skills like AI-assisted programming or graphic design. “If you don’t know how to use AI, you won’t even get an interview,” she said.

Targeting university students’ anxiety, many training centres have launched “AI crash courses” on campus, with fees starting at 10,000 RMB. Lu has seen the adverts several times but remains torn about enrolling: “It’s such a huge sum that I can’t bring myself to spend it. But if I don’t learn, I’m afraid I won’t be able to find a job.”

As the impact of AI on the job market becomes increasingly evident, Chinese authorities have responded intensively over the past month, sending out clear policy signals.

The government work report released in early March called for taking “more effective measures to facilitate employment and entrepreneurship through better adaptation to the development of AI technologies”. The subsequent 15th Five-Year Plan (2026-2030) further emphasised strengthening AI’s role in job creation to address the impact of new technologies like AI on employment.

During a public forum on 22 March, Han Wenxiu, executive deputy director of the Office of the Central Committee for Financial and Economic Affairs, pledged that amid AI’s ubiquitous impact, the authorities would respond comprehensively to promote high-quality and full employment.

Notably, in Chinese state media narratives, AI is more often portrayed as a “creator” rather than a “replacer” of jobs. Such reports usually cite a set of data released in a job report by the World Economic Forum: by 2030, the global workforce transformation will create 170 million new jobs while displacing 92 million, resulting in a net gain of 78 million positions.

However, another report released in February by Citrini Research presents a different scenario: while AI may create new jobs, each new position could make dozens of existing ones obsolete, while paying a fraction of what the old ones did.

Zhu Xufeng, a longtime researcher of technology governance and dean of Tsinghua University’s School of Public Policy and Management, told Lianhe Zaobao that from an individual perspective, technological substitution does indeed bring “short-term pain”; but viewed over a longer historical horizon, it is an inevitable pattern in human development spanning millennia.

He explained that as long as technology advances, jobs will be displaced, because the very essence of technology is to improve efficiency and reduce the need for labour. In this process, the replacement of existing roles is inevitable, and the labour thus released tends to flow into new industries and jobs.

Zhu added that China is facing a worsening ageing population and declining birth rates, with labour costs continuing to rise. In this context, firms turning to AI is in itself a rational choice. He said, “If technological development is restricted in order to preserve jobs, it will instead weaken the country’s overall competitiveness.”

Meanwhile, Shan Wei, a senior research fellow at the East Asian Institute (EAI) of the National University of Singapore (NUS), noted when interviewed that academics are still debating whether this round of AI advancement follows the trajectory of previous industrial revolutions or fundamentally replaces human labour and upends the existing economic system. 

He said that the crux of the debate is whether AI, as a new technology, can expand the scope of human activity. Otherwise, many of the roles and jobs people currently hold may eventually be replaced. But if AI can lead humanity into new domains, broadening our knowledge and range of activity, then “amid the vast expanse of the universe, humanity may yet rediscover a place to anchor its existence”.

He added, “It is just that we do not know for sure what these new jobs and tasks will be.”

When Li Li in Anhui was looking for a new job, she did not intentionally avoid the ones easily replaced by AI. She admitted that there were few jobs to begin with; even if they will ultimately be replaced by AI, she is not thinking that far ahead: “I can only focus on the present and try to earn a little more for now.”

She is well aware that she is competing for the limited human-held positions not only with other job seekers displaced by AI, but also with a constant stream of university graduates. Until a new employment equilibrium is formed, the upheaval and uncertainty brought about by AI will continue.

Amid anxiety about being replaced by AI, increasingly more working professionals are actively trying to build a “moat” for themselves.

Chinese language and literature graduate Zhu pushed himself to become the “tech guy” in his company. He went from using AI to handle writing tasks, to using it to write code, and most recently to deploying OpenClaw — afraid that he would be left behind at any turn.

He believes that keeping up with the latest in AI technology is one of the few certainties he can cling to. “If human work becomes about directing AI, then I would try my best to be a pioneer when the time comes,” he said.

For programmer Shi Wenyu, the strongest “moat” is likewise technical skill itself. Even if AI eventually replaces most programmers, he still felt that senior engineers in the top 10% of the field would have a place.

With the proliferation of vibe coding — where even those untrained in coding can get AI to write code — Shi has also started to see the value of “manually written” code. He commented that AI-generated code is mostly based on generic logic, whereas code that truly fits a specific need has a kind of “human aesthetic” that AI cannot replicate. “That’s something only experienced programmers can do,” he said.

In the legal profession, another field widely seen as highly susceptible to replacement by AI, Wang Jun, a lawyer at the Guangdong Weiqiang Law Firm, said when interviewed that AI can already handle a large volume of compliance checks on written documents and non-litigation work such as basic legal advice.

But he also pointed out at the same time that much of legal work is not just about the statutes themselves. “AI can only deal with the provisions of the law; it is unable to tackle the human and moral dimensions,” he shared.

On the subject of humanity’s “moat” in the age of AI, NUS’s Shan said that aesthetics, empathy and creativity remain areas that AI would struggle to truly reach in the near term.

He said that even when it came to directing AI agents to complete tasks, the results would differ from person to person, often because of idiosyncrasies in personal taste. Likewise, although there are now many AI chat tools offering companionship, AI still finds it hard to perceive the most delicate and genuine emotions in the human heart.

Tapping on the phrase often used to describe a doctor’s duty, “to cure sometimes, to treat often, to comfort always”, Shan said that the AI might score 90 out of 100 for the first part; 70 for the second part; but probably fail for the last part.

At the same forum where Anthropic’s Amodei made his prediction about a countdown for many professions, Demis Hassabis, CEO of Google’s AI firm DeepMind, raised a broader question:

In a “post‑scarcity world” where humanity has immense material wealth and where money might not even be necessary anymore, beyond the displacement of jobs by AI, can societies find ways to distribute the new productivity and wealth more fairly? And are existing institutions robust enough to support such a transformation?

Hassabis admitted that research on these questions is still far from sufficient. Given his prediction that AGI could be achieved within the next five to ten years, “the time we have left to think about this is already very tight.”

In China, AI adoption has surged over the past year. Official reports showed that by the end of 2025 there were 602 million AGI users. At the same time, differences in regional development levels and employment structures meant that the shocks and opportunities brought by AI would vary greatly across different groups.

Shan noted that the “digital divide” seen in the internet era may reappear in the age of AI. Those who master AI technologies would be the first to reap benefits, widening wealth gaps. Meanwhile, just as internet technologies eventually spread and narrowed the divide, the gap in AI use would also shrink as the technology spread.

He felt that judging from the US experience, the biggest AI impact falls on white-collar jobs in the service sector, such as programmers, analysts and customer service representatives. In China, the economy has a large manufacturing base and a relatively smaller services sector, so the short-term shock would be less severe than in the US. The deep question however, is whether China’s current social security system is ready for large-scale unemployment.

Shan said, “When the government channels more resources into ‘new quality productive forces’, if jobs and consumption that are closely linked to ordinary folk fail to keep pace, this structural gap might be an even bigger challenge we have to confront in the future.”

This article was first published in Lianhe Zaobao as “AI大军压境抢饭碗 打工族技荒心更慌”.

Related: The illusion of independent AI: How the US and China control the machines | A short-lived catch: China’s AI lobster frenzy