InDepth:HowAIIsRewiringWhite-CollarWorkinChina

“Did you know? My mom got laid off!”

Wu Qiong’s daughter leaned toward a classmate at kindergarten, eager to share the “news.”

The child did not understand the gravity of a layoff. To her, it was simply the latest novelty in her mother’s life.

For Wu, the moment carried a painful mix of heartache and dark humor. In April, her employer — a web-development firm that had pivoted to artificial intelligence a year earlier — slashed its workforce. Wu estimated that more than 60% of employees were laid off, while her 50-person department was eliminated entirely.

Wu had worked as an AI data analyst, tracking client token usage and modeling profit margins. But the company’s AI pivot had failed to turn a profit. Rising token costs, fixed client budgets and expanding feature requests squeezed margins.

What troubled Wu most was not only losing her job, but the way it happened.

“I knew AI was developing fast, but I didn’t expect it to be this fast,” she said.

At her boss’s request, Wu had used AI to automate her own highly repetitive data-reporting work. Once the logic was set, the system ran smoothly without human keystrokes. In building the automation, Wu had helped engineer her own obsolescence.

Concluding that she could not keep pace with the technology, Wu left the AI sector for a job in traditional manufacturing, taking a 30% pay cut. Still, she has no illusions about escaping the shift.

“Now it seems no industry can hide from AI,” she said.

Wu’s experience reflects a broader transformation sweeping China’s workforce. Over the past three years, rapid advances in large language models have turned AI from a theoretical threat into a daily worry for many workers. But companies’ drive to cut costs and raise efficiency is not producing only straightforward automation. It is also rewriting job descriptions, changing performance metrics and reshaping the labor market.

Doing more with less

For many workers, AI has not eliminated tasks so much as compressed them, increasing the burden on those who remain.

Yang Ru, a publicity employee at a music distribution company, used to rely on outsourced workers at partner streaming platforms to sort songs and categorize metadata. Recently, those platforms dismissed their contractors, believing AI could handle the work.

The immediate result for Yang was a sharp drop in efficiency. Inquiries that once received answers in five minutes now take half a day. She later learned that the outsourced workload had not disappeared; it had simply been shifted to the platforms’ remaining full-time employees.

Overwhelmed, those workers now rely on traffic data and tags to screen music quickly. That has made it harder for niche artists to secure homepage placements, drawing complaints from Yang’s clients.

“It’s like a butterfly effect,” Yang said. Platform executives believed AI could replace human labor, but the resulting bottlenecks eventually landed on intermediaries like her.

The same dynamic is reshaping pay and job scope.

Li Meng, a 35-year-old visual designer, was laid off from an internet healthcare company after eight years. Back on the job market, she found that employers were combining visual design, video editing, social media management and AI image generation into single roles — while cutting monthly salaries from about 10,000 yuan ($1,472) to between 5,000 yuan and 8,000 yuan.

Employers also expect faster turnaround. Projects that once took 10 days are now expected in five.

Zhu Zijian, an illustrator, felt a similar squeeze. AI sped up his output, prompting clients to demand shorter deadlines and lower fees. Even after adopting AI tools, Zhu’s overall income fell, and he was eventually laid off.

One person doing the work of three has become the industry norm, he said.

A micro-drama production crew films in Zhengzhou, Henan province, in September. Photo: Guan Cong / Caixin

The skill treadmill

AI is also changing how workers are evaluated.

Huang Xiao, a civil servant who monitors public opinion and writes reports, noticed that a struggling colleague had suddenly begun producing five or six high-quality reports a day with the help of AI. Even after management banned the practice, the colleague continued using AI to draft reports, then manually edited the results.

As volume became an easy-to-follow performance metric, Huang said, the premium placed on human experience and writing quality began to erode.

Professional boundaries are blurring as well.

Liu Yang, an intern on a tech company’s AI video project, found that product managers were now expected to learn AI coding tools and commit code directly. As AI enables copywriters to handle basic visual tasks, traditional qualifications are shifting.

“Now, if you don’t know how to code, you barely meet the threshold,” Liu said.

The result is persistent anxiety over skills. To stay competitive, workers must master capabilities far beyond their original job descriptions.

When Xia Xue’s company shifted from live-action micro-dramas to AI-generated ones, it dismissed its entire Beijing operations team. Xia had already moved into AI video editing, but she was laid off anyway. The lower technical barrier allowed the company to hire cheaper, less-experienced replacements.

Reshaping the market

Many workers acknowledge that AI is not the sole driver of layoffs. Falling corporate revenue and China’s broader economic slowdown remain the primary pressures. AI gives strained companies another option: maintaining output with fewer people.

Recruitment data point to the shift. According to Zhaopin, a major Chinese employment platform, hiring demand in the first quarter of 2026 fell 29% year-on-year for editing roles, 23% for customer service positions and 21% for visual design jobs. By contrast, demand for roles requiring large language model skills rose 73%.

A 2025 Zhaopin report found that 78.2% of professionals use AI weekly, while nearly half said they had been explicitly required to upgrade their AI skills over the previous year.

Zhaopin Vice President Li Qiang said that does not mean traditional roles will disappear entirely. In translation, for example, recruitment demand fell sharply in 2024 but began recovering by 2026. The reason, he said, is that requirements have changed. Companies now want candidates who can translate, negotiate, receive clients and manage cross-cultural communication.

“It is actually merging several positions into one,” Li said.

New roles are emerging at the same time. According to Zhaopin, recruitment demand for AI engineers rose 17% year-on-year in the first quarter. Demand for basic data roles such as AI trainers also increased 17%, while demand for AI product managers surged 81%. Hiring is expanding beyond major technology companies into manufacturing, healthcare and services.

Li predicted that companies may eventually keep only 40% of their core staff as full-time employees, relying more heavily on flexible labor.

“Building an organization will be like playing with Legos — you move pieces wherever they are needed,” he said.

With technological support, one person can perform the work of several, potentially accelerating the rise of small teams that resemble “one-person companies.”

Calls for policy action

The macroeconomic impact of AI-driven labor shifts is drawing increasing scrutiny.

At a recent seminar, Cai Fang, a member of the Chinese Academy of Social Sciences, warned that AI would hit young and older workers especially hard. Young workers’ entry-level skills are being devalued, while older workers face a widening digital divide.

Huang Tiejun, chairman of the Beijing Academy of Artificial Intelligence, has forecast that artificial general intelligence (AGI) could arrive by 2045. Cai said that in a theoretical AGI era, all jobs could potentially be replaced.

“Therefore, throughout this process, we must continuously address the growing structural employment contradictions, and policy efforts must be consistently strengthened,” Cai said.

He advocated a “U-shaped” human-capital model that invests heavily in both early childhood education and lifelong retraining.

Lawmakers are also proposing guardrails.

Ma Yide, a National People’s Congress deputy and academic, has suggested creating an AI employment-impact assessment mechanism. Similar to environmental reviews, large companies deploying AI would be required to submit proactive reports detailing which roles would be affected and what retraining plans would be offered.

Ma also called for updating China’s social-security system to recognize “technological unemployment,” providing affected workers with extended benefits and training subsidies.

Ma told Caixin that an employment-impact assessment mechanism would serve as an early-warning and response system, helping policymakers stay alert to labor-market shifts without restricting technological progress.

“Doing this isn’t about erecting hurdles for technological development, but making the distribution of tech dividends and social costs more transparent and fair,” he said.

Wu Qiong, Yang Ru, Li Meng, Zhu Zijian, Huang Xiao, Liu Yang and Xia Xue are pseudonyms.

Contact editor Han Wei (weihan@caixin.com)