China’s AI competition is beginning to change jobs far from the model labs and data centers that usually dominate the debate. South China Morning Post reports that a Beijing research analyst, Clare Zhang, lost her position after her corporate-advisory employer concluded that AI had reduced the number of analysts it needed.
Zhang’s case is a detailed account, not a national employment survey. But it shows how automation can reach work that once depended on people reading company filings, assembling public records, writing market studies, and preparing charts. Her former firm produced reports that could run to 500 pages, according to SCMP. The company said AI could gather and organize information, analyze data, and generate charts more quickly than its human team.
The report is also a reminder that adoption is not the same as a clean productivity gain. Zhang said her former team warned clients that AI could invent sources and produce polished but unreliable findings. Yet clients increasingly chose to bring research in-house with AI because outside reports were seen as too expensive. The new pressure is therefore not only whether AI can perform a task, but whether customers accept a lower-cost output despite its risks.
Corporate Research Is Becoming an Early Test Case
SCMP reported that Zhang had worked for four years as a research analyst at the Beijing firm until late 2025. Her job involved reviewing filings and public records to assess companies and competitors for projects heading toward listings. The work was data-heavy and structured, which makes it easier for AI tools to assist with parts of the process.
Her human-resources department explicitly cited AI when it dismissed her, she told SCMP, saying the firm could no longer justify as many analysts. Zhang said two former colleagues were laid off soon after. She later learned that her former manager was also struggling with fewer clients as companies shifted research work inside their own organizations.
This is a narrower picture than the claim that AI is shrinking every white-collar profession. Analysts, editors, call-center staff, accountants, lawyers, and software developers have different workflows and regulatory obligations. Yet the research-advisory example helps explain where pressure may appear first: jobs that involve collecting large amounts of structured public information and converting it into standardized presentations for clients.
EastFrontier reported in June that China’s tech workers feared being optimized out as AI agents arrived in the office. The Zhang case gives that anxiety a concrete setting. Her employer was not replacing every form of human judgment. It was reducing headcount around research processes where clients were willing to accept a cheaper, faster automated alternative.
The potential savings may be real, but so are the quality-control questions. A market study can influence an investment, a product decision, or a listing plan. If an AI system invents a source, misreads a filing, or fails to distinguish a forecast from a confirmed result, the economic cost can exceed the savings from a smaller team. Companies using AI to bring work in-house still need people who can verify sources, assess uncertainty, and take responsibility for the final recommendation.
Chinese Courts Have Drawn a Limit Around Cheaper AI Labor
SCMP’s report places Zhang’s experience alongside a May court case in Hangzhou. A fintech firm had dismissed a 35-year-old quality-inspection employee after he refused a demotion and pay cut. The company said an AI replacement would be cheaper. The court ruled that the dismissal was unlawful, according to the SCMP article.
The two cases are not the same. Zhang’s report describes a corporate-advisory employer reducing its analyst team as clients brought work in-house. The Hangzhou judgment involved a different company, a different role, and a specific employment dispute. Still, the court decision establishes that the desire to cut costs through AI does not automatically make a dismissal lawful.
That legal point is likely to matter as companies experiment with AI-enabled reorganizations. Employers may be able to change workflows, create new roles, or reduce certain tasks. But they still have to comply with labor rules governing contracts, performance, transfers, and dismissals. Technology does not remove the requirement to follow employment law.
China’s policymakers are also trying to manage the broader transition. EastFrontier covered a State Council employment-first plan that targeted AI job creation in June. That policy orientation recognizes a tension at the center of AI adoption: the same tools that can reduce labor needs in one office may create demand for people who build, deploy, maintain, audit, and govern AI systems elsewhere.
The hard question is timing. Workers displaced from a research team may not immediately move into AI-related jobs. New roles can require technical skills, industry knowledge, or access to training that is not evenly available. Zhang’s experience underscores the risk that a company’s efficiency gains may arrive faster than a worker’s path into a replacement role.
AI Adoption Changes the Value of Human Review
The corporate-advisory case also suggests that the most durable human contribution may shift. AI can search public records and produce first drafts at high speed. It is less clear that it can reliably determine which sources are trustworthy, how a regulatory change affects a specific client, or whether a conclusion is sufficiently supported to guide an important decision.
Zhang’s former team warned clients that AI could generate convincing but unreliable material. That is not an argument against using AI. It is an argument for designing workflows in which systems accelerate repetitive work while people verify evidence, detect gaps, and explain uncertainty. A 500-page report may take fewer hours to produce, but a shorter report with unverified assumptions can still create a serious problem.
Employers will face choices about whether to use savings from automation to reduce staff, retrain workers, or shift employees into more complex review roles. Those choices will vary by sector. A small advisory firm facing price pressure may have less room to carry a large team than a bank or manufacturer that can redeploy staff across many functions.
The September report does not prove an economy-wide collapse in white-collar employment. It does show that China’s AI race is now affecting the business model of knowledge work. Firms are testing whether models can replace parts of research, and clients are testing whether the lower price is worth the trade-off in reliability.
For workers, the immediate issue is not an abstract debate about whether AI will create more jobs in the long run. It is whether employers give them a credible route to new roles as existing tasks are automated. For companies, the issue is whether a faster, cheaper report can meet the standard of evidence their customers and regulators require. Zhang’s account shows that both questions are moving from theory into workplace decisions.