Artificial intelligence is delivering substantial productivity gains in Chinese enterprises, according to new ILO research examining how firms are adopting the technology. But the findings also point to concerns over skills gaps, job displacement and the impact of AI on workers' future incomes.
The new ILO Research Brief, Artificial Intelligence Adoption in Chinese Enterprises: Productivity Effects, Workforce Implications, and Policy Challenges, was prepared by Ekkehard Ernst of the ILO Research and Satistics Department together with researchers at Renmin University of China.
The research draws on in-depth interviews with 21 enterprises and a survey of 1,591 professionals. The firms span manufacturing, finance, business services, construction, education, media and travel, and range from an eight-person start-up to a conglomerate employing 270,000 people. Every firm interviewed was either already using AI or had concrete plans to adopt it.
Among the professionals surveyed, 56 per cent see AI adoption as an inevitable trend, while 47 per cent believe AI creates more jobs than it displaces - a more optimistic outlook than surveys in many OECD countries have found. At the same time, 39 per cent expect AI to lead to declines in their income.
Significant gains, but difficult to measure
Firms that track AI's impact report substantial productivity improvements. At one insurance company, 300 customer-service employees increased the number of issues handled each day from 6,000 to 15,000. At a large insurance group, recruitment cycle times fell from 30 days to 13 days, a 57 per cent reduction. A smart manufacturing facility reported a 30 per cent increase in production efficiency.
Yet such measurement remains the exception rather than the rule. Most firms studied lack systematic frameworks for assessing AI's impact, particularly beyond conventional productivity measures. The brief highlights the need to better capture changes in job quality, working conditions and the composition of workers' tasks.
The productivity figures reported by firms are self-reported and have not been independently verified. The study also notes that the 21 enterprises interviewed were purposively selected, meaning the firm-level findings cannot be statistically generalized to all Chinese enterprises.
Work is changing alongside technology
The gains reported by firms are concentrated in repetitive and data-intensive tasks such as document processing, customer-query handling, résumé screening and data collection. The brief finds that displacement pressures could therefore be particularly significant for routine clerical, administrative and customer-service roles.
Firms also report challenges related to employee resistance, skills, AI output quality, data security, regulation and integration with existing systems. One ed-tech company identified a marked age divide in AI capabilities among employees over 40, while other firms highlighted broader gaps in AI literacy and the ability to use new tools effectively.
Despite these pressures, the research finds that the predominant model remains one of human-AI collaboration rather than full automation. Across the enterprises studied, AI is generally being incorporated into hybrid workflows in which human judgement and oversight remain important.
"Chinese firms are not replacing workers with AI so much as reorganising work around hybrid human-AI workflows," said Ekkehard Ernst. "The real bottleneck is not the technology but the skills and management capacity to use it well. That is why policies for lifelong learning, transition support and accessible AI services for small and medium-sized enterprises will matter more than the pace of technological progress itself."
Skills, transitions and inclusive adoption
The brief points to four areas for policy action: strengthening AI skills and lifelong learning, particularly for mid-career and older workers; supporting workers' transition towards higher-value tasks; developing better frameworks for measuring AI's effects on productivity, job quality and working conditions; and helping smaller firms access AI through shared platforms, training and affordable services.
These measures, the brief argues, will be important to ensure that productivity gains from AI are accompanied by decent work and are distributed more broadly across workers and enterprises.
