AI applications have significant implications for global workers. The key trends from China’s implementation of AI tools could provide important lessons for the wider world.
Market-driven development with little regulatory oversight
In China, AI solutions for work have been largely driven by market forces, with little regulatory oversight or input from workers. New products and services are often directly released to the market as soon as they become technologically feasible. To secure a foothold in the workplace solution sector, it is common for Chinese providers to offer customized functions tailored to the specific needs and preferences of their clients. This approach has led to the aggressive design and hasty rollout of AI tools, where firms prioritize profits over the well-being of workers.
While AI indeed improves overall productivity by constantly pushing up the benchmarks, it often overlooks the circumstances under which better results are achieved. As mentioned earlier, in the case of food delivery drivers, the AI system went ahead to reduce target delivery times for all, ignoring the fact that the new goals can only be attained by breaking traffic rules. As a result, productivity enhancement facilitated by AI is often obtained at the expense of the workers, with firms facing few legal consequences.
At present, there is not a specific law in China to regulate the deployment of AI in workplaces. The government largely takes a hands-off approach to the practices of firms that utilize AI, intervening only when some extreme cases gain public and media attention.
After a Shenzhen gaming firm was reported to have installed AI-enabled surveillance cameras on each employee’s workstation in early 2022, the China Youth Daily – a mouthpiece of the Communist Youth League of China – criticized the practice in an editorial. The firm subsequently removed the cameras but faced no further punishment. In fact, such mild rebukes are hardly likely deter Chinese firms from jeopardizing the well-being of their employees when tempted by the substantial benefits from adopting AI for work.
Blurring of personal and professional spheres
The significance attached to work in Chinese culture, compared to Western societies, has led to a higher tolerance among people when their personal space is eroded by work matters. In China, there is a deeply ingrained culture in which people value hard work and often prioritize work above personal leisure. The country only introduced the concept of a two-day weekend in 1995.
AI tools are further blurring the already vague boundaries between work and life in China by extending the reach of firms beyond traditional office hours and workspaces. Functions include constant location tracking, online activity monitoring and intrusive notifications. Workers are frequently expected to be prepared to work on demand, turning what should be contingency arrangements into everyday occurrences. Through AI tools, employers gain greater access to workers’ personal data beyond what is necessary for the job. This is exacerbated by pervasive data collection and a comparatively utilitarian view of data privacy in China (compared to its Western counterparts).
It is worth noting that major developers of AI work solutions in China also own a range of other applications used in people’s daily lives. Alibaba, Tencent and ByteDance – developers behind China’s top three pieces of collaborative work software – also operate popular leisure and utility applications ranging from payment and ecommerce to video streaming and online education. This dual role affords them the capability for extensive worker profiling through the centralization of data collected from both professional and personal channels.
Intensified competition and diminishing returns
Both white-collar professionals and gig economy workers on platforms noted an intensified level of competition due to AI. Employers, facilitated by AI, have access to a broader talent pool in which candidates need to meet higher standards to land a job. AI’s capability to quantify employee activities and behaviours also makes performance evaluation more challenging for workers, as they can now be compared with their peers on more metrics to discern their value to the firm. This heightened competition has led to the intensification of work tasks, as workers must adapt to both the ever-accelerating pace set by AI and the additional efforts required to stay competitive in an AI-dominated job market.
While AI algorithms are driving workers to extremes, the benefits of this increased productivity are not being equitably shared. Instead, many workers interviewed for this research paper have seen declining economic returns and worsening work conditions over time. Business models enabled by AI and the designs of algorithms have contributed to this phenomenon.
While AI algorithms are driving workers to extremes, the benefits of this increased productivity are not being equitably shared.
Gig economy platforms typically adopt a ‘bait-and-switch’ strategy, where firms initially offer attractive incentives to draw in workers or customers, only to reduce or remove these benefits once the platform has secured a dominant market position. In China, platforms like Didi and Meituan initially attracted workers and customers by offering substantial subsidies per order during their initial expansion phase. The high income enticed many individuals to join these platforms as workers, with some even quitting their full-time jobs. However, both platforms have significantly cut the subsidies to workers in recent years as the companies became dominant players in their respective fields, which has affected worker pay. All five drivers for ride-hailing apps interviewed for this research paper said that they have to accept more orders and work for longer to make the same amount of money that they earned a few years ago.
In addition, algorithm design also contributes to workers’ intensified workloads and diminished returns. Researchers highlight the role of ‘gamification’ in algorithm design, wherein tactics usually seen in video games are employed to shape worker response to maximize the platform’s profits. In ride-hailing apps, for instance, various strategies are implemented to encourage drivers to stay active for extended periods. Drivers on the Didi app said the platform will reward drivers who are available during peak times with more lucrative orders, which prompts many to start their day as early as 7 a.m. and then not to log off until 10 p.m. to capture both morning and evening peak times. To further incentivize drivers, the platform offers bonus rewards if a driver’s daily earnings hit a certain mark. However, several drivers interviewed pointed out that as their earnings approach these thresholds, the system started to allocate them low-value rides, meaning they needed to complete a higher number of rides and stay online for longer to hit the target. As one driver noted, ‘Sometimes, it’s just not worth it.’
Diverging AI strategies along industry value chains
The research for this paper found that in China, smaller firms and those lacking a competitive edge tend to be more aggressive in their use of AI workforce management compared to industry leading firms. This pattern echoes the perception that low-skilled jobs are more vulnerable to the manipulation and exploitation of AI, as firms that employ these roles often operate in the lower end of industry value chains, such as in primary production.
These companies, such as manufacturers or telemarketing firms, often face stiff rivalry from competing suppliers due to the nature of their products or services compared to those offered by higher-end firms, such as tech companies. Additionally, these companies are sensitive to price fluctuations since they operate on much narrower profit margins. Given the nature of their businesses, these firms have limited opportunities to increase revenue or reduce costs in areas like rent and raw materials. As a result, maximizing the value extracted from labour becomes essential for maintaining their competitiveness in the market. In this context, AI emerges as a valuable tool for controlling the labour process and enhancing productivity. However, companies with limited resources often adopt basic AI solutions focusing on worker monitoring, instead of optimizing operations through big data analytics, which demands greater investment and maintenance.
By contrast, firms sitting at the top of the industry value chain, especially tech giants, typically use more advanced AI systems and big data analytics to identify new trends, make forecasts and update datasets promptly to fine-tune algorithms. Unlike smaller, less-known firms, they are more mindful of potential legal and reputational impacts. For these top-tier firms, although extended hours and increased workloads feature in their work patterns they do not necessarily equate to higher profits; instead, these firms rely more on the innovative contributions of their skilled workforce for growth. In addition, considering the generally higher educational background of employees in these firms and their stronger position in the job market, employers are cautious about implementing overly intrusive control measures that might repel talented workers.
While the use of AI systems for higher skilled workers may be less immediately noticeable or disruptive to employees on the surface, AI tools can still be powerful or even detrimental. The comprehensive control and monitoring of work processes through AI enables firms to utilize all sorts of data against workers, especially when disputes arise.
These findings align with the technological divide highlighted in global value chain literature. This disparity in the capacity of firms and countries to utilize AI will likely intensify the polarization within global value chains. Leading firms, like Google and Amazon, primarily based in the Global North, are more capable of leveraging AI to its full potential and incorporating more nuanced considerations of human rights and worker well-being. Firms engaged in low value-added production segments are predominantly situated in the Global South. Limited by their technological capabilities and resources, these firms might only resort to less sophisticated AI tools that focus on enhancing productivity. Furthermore, less developed countries tend to lack robust legal protections and social awareness on privacy and labour rights. It can potentially pave the way for AI to be used in ways that exploit workers and exacerbate conditions for an already vulnerable workforce.
Counterproductive and unintended results
While AI promises greater efficiency, its initial deployment often demands substantial human input. This requirement frequently leads to additional workload and, in many instances, it can slow down operations.
The training of AI algorithms requires large amounts of data, and the initial data processing is still manually done by human workers. Some firms hire low-paid workers in developing countries to label and clean raw datasets to facilitate AI training. In China, this work is often done in-house. For instance, as part of what was originally planned as a week-long project, a software engineer from Shenzhen mentioned that their team had to dedicate an additional two days solely for generating and inputting the relevant data for AI analysis. The additional data processing work added strain to the already gruelling ‘996 schedule’ (9 a.m. to 9 p.m., six days a week) of Chinese tech workers. HR professionals face similar problems. Some mentioned that they spend additional hours collecting and digitalizing employee data to train AI models for candidate matching, as the information is often not in AI-ready formats.
There are heightened risks for firms becoming overly reliant on AI for work management. In addition to the increased data security risks, glitches of AI systems can lead to major operational disruptions and business losses. This over-dependence on AI might start a detrimental cycle in which firms, having replaced human roles with AI, find the remaining workforce increasingly leaning on AI for decision-making due to having a limited number of workers. Such a scenario could cause firms to become short-sighted and less agile in rectifying mistakes.
Many workers develop strategies that aim to please or resist AI algorithms, which could hinder the overall objectives of the firms. For example, AI evaluation prompts some workers to prioritize tasks that can be easily recognized by the system, over more exploratory work that could benefit the firm’s strategic interests in the long term. Some gig economy workers also devise strategies to ‘game’ the system. For example, some taxi drivers interviewed for this paper resorted to ‘cheat’ software – often a mobile app plugin – to capture high-value fares, while food delivery drivers colluded with restaurants to generate fake orders and then shared the platform bonuses between them. Employees report lower job satisfaction and increased mental stress in environments where aggressive AI systems are deployed, leading to elevated employee turnover rates that could increase recruitment costs.
While many companies in China see AI as a linchpin to their business success, it does not guarantee greater benefits, especially in the short term. In fact, as shown by the examples above, the substantial investment – both from procurement and subsequent training of algorithms and staff – does not always correlate with clear business returns and could further strain the financial resources of firms.
