In China’s IT industry, controversy is growing as more companies demand that employees train artificial intelligence (AI) agents to carry out their own work.

MIT Technology Review reported on April 20 that some companies are pressuring employees to document work procedures and decision-making methods and transfer them to AI tools.

At the center of the dispute is a GitHub project called Colleague Skill. The tool claims it can distill a specific co-worker’s work style and tendencies and recreate them as an AI agent. Its developer, Tianyi Zhou (저우톈이), an engineer at the Shanghai AI Laboratory, said the project began as satire. He explained that it reflects rising fears of layoffs as AI is adopted and a stronger trend of companies requiring workers to automate themselves.

The way the tool works is symbolic. When a user selects a colleague to copy, it pulls chat logs and files from Chinese enterprise collaboration apps Lark and DingTalk and compiles that person’s work manual and traits. It is said to incorporate elements such as speaking style and how the person responds, going beyond a simple task manual, which is reported to have caused a strong reaction in workplaces. A tech worker in Shanghai said, "I reproduced a former colleague and it was surprisingly similar," adding, "It even helped with code debugging, but it also felt creepy."

After OpenCloud became a trend in China, workers on the ground said management became more inclined to push agent experiments on employees. AI agents can operate computers, read and summarise news, reply to emails and handle booking tasks, but there are still significant limits in real corporate environments.

In this situation, companies are said to be requiring employees to document detailed tasks in order to expand the use of AI agents. Experts analyse that this can help companies accumulate work data and decision-making patterns and use it to judge which tasks can be automated.

Reactions on the ground are mixed. Some engineers say they feel anxious that their work is being broken down into data units and becoming more replaceable. Others say there are still major technical limits to AI fully replacing humans.

In response, "anti-distillation" tools have also emerged. A Chinese AI product manager 공개한 a tool that deliberately makes work documentation vague to reduce the usefulness of AI training. The video drew attention after receiving millions of reactions on social media.

As AI adoption accelerates, the boundary between improving work efficiency and turning individual labour into data is emerging as a new point of conflict in China’s IT industry.