In China, tech employees are being directed by their supervisors to prepare AI agents to take their positions—leading to a period of introspection among previously excited early adopters.
Earlier this month, a GitHub initiative known as Colleague Skill, claimed that employees could use it to “distill” the skills and characteristics of their colleagues and recreate them using an AI agent, gained significant traction on Chinese social media. Although the project was intended as a parody, it resonated with tech professionals, many of whom informed MIT Technology Review that their employers are urging them to log their workflows to automate certain tasks and processes using AI agent tools such as OpenClaw or Claude Code.
To initiate Colleague Skill, a user designates the coworker whose responsibilities they wish to mimic and inputs basic profile information. The tool subsequently imports chat histories and documents from Lark and DingTalk, two widely-used workplace applications in China, generating reusable guides detailing that coworker’s tasks—and even their distinctive quirks—for an AI agent to duplicate.
Colleague Skill was developed by Tianyi Zhou, an engineer at the Shanghai Artificial Intelligence Laboratory. Earlier this week, he mentioned to the Chinese publication Southern Metropolis Daily that the project began as a publicity stunt, driven by AI-related job cuts and the increasing trend of companies requesting employees to automate their own roles. He did not reply to further requests for comment.
Internet users have found humor in the concept behind the tool, making light of automating their colleagues before themselves. Nevertheless, the popularity of Colleague Skill has ignited extensive discussions regarding workers’ dignity and individuality in the era of AI.
After encountering Colleague Skill on social platforms, Amber Li, 27, a tech professional in Shanghai, utilized it to replicate a past coworker for a personal experiment. Within moments, the tool produced a document detailing how that individual performed their job. “It is surprisingly good,” Li remarks. “It even captures the person’s little quirks, like how they respond and their punctuation habits.” With this skill, Li can employ an AI agent as a new “coworker” to assist her with debugging her code and providing instant replies. It felt eerie and unsettling, according to Li.
Nonetheless, substituting coworkers with agents may become a commonplace practice. Following the rise of OpenClaw as a national sensation, employers in China have been urging tech workers to experiment with agents.
While AI agents can manage your computer, analyze and summarize news, respond to emails, and make restaurant reservations on your behalf, tech workers report that their practical use has, to this point, been limited in professional settings. Asking employees to create manuals outlining the intricacies of their daily tasks, as Colleague Skill does, is one approach to help close that gap.
Hancheng Cao, an assistant professor at Emory University specializing in AI and labor, posits that companies have justifiable motives for encouraging employees to develop work guides like these, beyond merely following a trend. “Firms gain not only internal experience with the tools, but also richer data on employee expertise, workflows, and decision-making patterns. This enables companies to identify which aspects of work can be systematized or codified and which still rely on human judgment,” he explains.
However, to employees, creating agents or even blueprints for them can feel odd and isolating. One software engineer, who spoke with MIT Technology Review anonymously due to concerns over job security, trained an AI (not Colleague Skill) on their workflow and found the process felt reductive—as if their work had been condensed into modules in a manner that made them simpler to replace. On social media, workers have resorted to dark humor to convey similar sentiments. In one comment on Rednote, a user noted that “a cold farewell can be turned into warm tokens,” humorously suggesting that if they use Colleague Skill to distill their coworkers into tasks first, they themselves might endure a bit longer.
The push for creating agents has also inspired inventive counteractions. Frustrated by the notion of reducing an individual to a skill, Koki Xu, 26, an AI product manager in Beijing, released an “anti-distillation” skill on GitHub on April 4. The tool, which Xu constructed in about an hour, aims to frustrate the process of establishing workflows for agents. Users can select from light, medium, or heavy sabotage modes based on how closely their supervisor is monitoring the process, and the agent reformulates the content into generic, non-actionable language that would yield a less effective AI replacement. A video Xu shared about the project became viral, amassing over 5 million likes across platforms.
Xu informed MIT Technology Review that she has been tracking the Colleague Skill phenomenon from the beginning and that it has prompted her to reflect on alienation, disempowerment, and the wider implications for labor. “I initially considered writing an op-ed, but decided it would be more impactful to create something that counters it,” she states.
Xu, who holds both undergraduate and master’s degrees in law, mentioned that the trend also raises legal issues. While a company might argue that work chat histories and materials produced on a work laptop are corporate assets, a skill like this can also capture aspects of personality, tone, and judgment, making ownership much less clear. She expressed hope that Colleague Skill encourages broader discussions about how to safeguard workers’ dignity and identity in the AI era. “I believe it’s essential to stay informed on these trends so we (employees) can engage in shaping how they are utilized,” she asserts. Xu herself is an enthusiastic AI user, with seven OpenClaw agents established across her personal and work devices.
Li, the tech employee in Shanghai, mentions that her company has not yet developed a method to replace actual workers with AI tools, primarily due to their unreliability and the requirement for constant supervision. “I don’t feel like my job is immediately at risk,” she states. “But I do feel that my value is being degraded, and I’m unsure how to address it.”
