⚡ Quick Summary
Published by Chatham House, the Royal Institute of International Affairs, this research paper by Nikki Sun examines how AI tools are deployed in Chinese workplaces and what that means for workers. It draws on more than 50 interviews conducted during two months of fieldwork in Shenzhen and Beijing in early 2023, with executives, developers, workers, HR professionals, scholars and lawyers, and on a review of over 20 AI products used by Chinese employers.
The central argument is that AI integration in China is largely driven by market forces and competition, favouring business interests over workers, and that the balance of power is shifting further towards employers who control data and algorithms. The paper identifies four mechanisms behind the tools — workflow automation, datafication, big data analytics and platformization — and reports that efficiency gains have not translated into better job quality, with higher workloads, stress and job insecurity.
It closes with recommendations under two headings: worker empowerment, covering protections in AI and data regulation, worker representation, transparency and data rights, and upskilling; and checking employer power, covering monopolies, limits on data collection and control of the pace of deployment.
🧩 What’s Covered
The paper moves in order from context and method to the employment lifecycle, findings and recommendations.
- Introduction and methodology: the qualitative approach, more than 50 interviews in Shenzhen and Beijing in early 2023 with executives, developers, HR professionals, workers, scholars and lawyers, a review of over 20 AI products, and the analytical frame of shifting power between business owners and workers.
- Workplace AI in China and beyond: why adoption is fast, citing over 4,300 AI firms, a workforce of 782 million people and 540 million employees working online, plus data abundance, lax privacy laws, market competition and state support.
- From hiring to firing: four functions of workplace AI — workflow automation, datafication, big data analytics and platformization — and Table 1 mapping tools against recruitment, management, evaluation and personnel changes.
- Recruitment: AI-recommended job ads, automated CV screening, AI video interviews that analyse facial expression and gesture, background checks using facial recognition against police databases, data leaks, and bias; Table 2 lists a six-step application process for graduates.
- Management: algorithmic task allocation on Didi and Meituan, collaborative platforms DingTalk, Feishu and WeCom with over 300 million monthly active users, account suspensions, 'smile recognition' entry systems and reduced autonomy.
- Evaluation and personnel changes: people analytics, online activity monitoring, wearable devices and cameras, increased workload and opaque criteria, and the information asymmetry that shapes dismissals and labour disputes.
- Key findings and AI problems: market-driven development with little oversight, blurring of personal and professional spheres, diverging AI strategies along value chains, counterproductive results, top-down decision-making and worker adaptation outpaced by deployment.
- Conclusion and recommendations: worker protections in AI and data regulation, worker representation, data rights, upskilling, curbing developer monopolies, limits on data collection and control of deployment pace.
💡 Why it matters?
For anyone governing, procuring or auditing workplace AI, the paper documents what algorithmic management looks like in practice: candidates scored by video, screen activity monitored, performance metrics that drive longer hours, and collected data used in dismissal disputes. It also shows risks for employers, including data security exposure, operational dependence on AI and reduced innovation. The recommendations connect to regimes the paper itself cites, notably the EU Artificial Intelligence Act's treatment of employment as a 'high-risk' area and China's 2022 Provisions on the Management of Algorithmic Recommendations, which require platforms to protect workers' rights.
❓ What’s Missing
The paper states explicitly that it does not aim to comprehensively evaluate the effectiveness or accuracy of AI solutions, map the specific technologies applied, or compare practices between China and other countries. Much of the evidence is dated: fieldwork in early 2023, platform data from 2024, and a youth unemployment figure of over 20 per cent from June 2023 before publication stopped. The account rests on two months in two cities. There is no assessment template or checklist, and how Beijing will weigh political risk against economic growth is left as a question for further research.
👥 Best For
Best suited to policymakers and labour regulators drafting AI or data rules that touch employment, to compliance and HR teams assessing algorithmic hiring and monitoring tools, and to researchers, journalists and civil society organisations seeking grounded evidence on algorithmic management in China. Also useful for global value chain and responsible sourcing analysts weighing worker risk in supplier firms.
📄 Source Details
Workplace AI in China: The changing profile of work and labour, a Research Paper by Nikki Sun, published by Chatham House, the Royal Institute of International Affairs, London, July 2024. Thirty-eight numbered pages of text; 42 PDF pages including front matter and imprint. ISBN 978 1 78413 615 4; DOI 10.55317/9781784136154. The imprint gives the citation as: Sun, N. (2024), Research Paper, London: Royal Institute of International Affairs, https://doi.org/10.55317/9781784136154. The extracted text covered all 42 pages.
