Chinese Tech Workers Are Training AI Doubles — And Their Bosses Are Requiring It
MIT Technology Review is reporting on a growing trend among Chinese tech workers: documenting their own workflows in order to train AI agents to replicate their job functions — and doing so at their employers' explicit request.
The phenomenon started as a social media joke — a "spoof" project — but struck a nerve. Multiple tech workers told MIT TR that management is now actively encouraging them to document processes so AI agent tools like OpenClaw and Claude Code can automate specific tasks and workflows.
What "AI Doubles" Means in Practice
An AI double isn't a chatbot that answers questions about you. It's an operational agent that can execute your actual work — handle emails, draft responses in your style, run reports, manage recurring tasks — with enough fidelity that your output continues even when you're not at your desk.
The OpenClaw setup for this is well-suited to the concept:
- SOUL.md — captures persona, tone, constraints, and decision-making style
- MEMORY.md — persistent context about preferences, relationships, recurring decisions
- AGENTS.md — defines what the agent can and can't do operationally
- Skills — specialized capabilities for specific recurring tasks
- Heartbeat — runs scheduled work autonomously, even while you sleep
The more accurately those files reflect how you actually work, the more effectively your agent functions as a genuine operational double.
The Employer Angle
This is the part that's creating anxiety among tech workers. When a company asks employees to document their workflows for AI automation purposes, the implicit question is: are they documenting their job so the AI can assist them — or so the AI can replace them?
The workers MIT Technology Review spoke to said the framing from management is "efficiency" and "continuity." Your AI double handles routine work so you can focus on higher-value tasks. Whether that framing holds long-term is an open question.
What's notable from an OpenClaw perspective: this is the first significant reporting of employers mandating AI agent adoption at a process level, not just offering it as a perk. That signals a shift from opt-in experimentation to operational expectation — at least in parts of the Chinese tech industry.
Why This Trend Will Reach the West
The pattern playing out in Chinese tech offices isn't culturally unique. It reflects a straightforward business logic: if an employee's routine tasks can be automated to run 24/7 without a salary, eventually some version of that calculation gets made everywhere.
The workers who fare best in this environment are the ones who:
- Build their own AI doubles first — owning the tool and workflow rather than having one imposed on them
- Use AI to work at higher leverage — freeing themselves from routine work to do the things agents genuinely can't
- Control their own infrastructure — a self-hosted OpenClaw setup means your AI double works for you, not your employer's IT department
The Self-Hosted Advantage
There's an important distinction between an employer-controlled AI agent and your own self-hosted one. A company-deployed agent running on corporate infrastructure is, effectively, owned by the company. The workflows it learns, the data it processes, the decisions it assists with — all of that stays on employer-controlled systems.
A self-hosted OpenClaw agent — running on your own hardware, with your own API keys, configured the way you choose — is yours. The distinction matters more as AI doubles become operationally significant rather than just convenient.
The Chinese tech workers building their own agents using OpenClaw and similar tools, rather than relying on employer-provided platforms, are making a bet on personal ownership of their own productivity infrastructure. That bet looks better every year.