China’s AI Agents Cross the Chasm From Tech to Profit
In a Beijing publishing office, 38-year-old manager Mr. Chen pays over 100 yuan (about $14) a month for AI agent services. In just weeks, these digital assistants have become indispensable — generating work presentations, crafting personalized English lessons for his child, and even producing animated explanations of scientific concepts on demand. He is not alone. Across China, a quiet revolution is underway as AI agents make the leap from “technically feasible” to “commercially profitable.”
According to a featured report by People’s Daily, Chinese AI agents are accelerating real-world deployment across multiple industries, marking a critical milestone in the maturation of the country’s artificial intelligence sector. The shift represents what industry insiders describe as a transition from experimental technology to a genuine economic force.
From ‘Vague and Elusive’ to ‘Gets Work Done’
The journey has been remarkably rapid. Zhang Yutao, co-founder of Moonshot AI, maps the evolution succinctly: in 2023, users were still figuring out “how to ask questions clearly.” By 2024, they had progressed to providing enough information for agents to understand complex tasks. Now, in 2026, the industry has entered a phase of “how to let it work independently and even fix its own mistakes.”
This transformation was catalyzed in no small part by OpenClaw (nicknamed “Lobster”), a free, open-source AI agent that became a cultural phenomenon in early 2026. Zhou Hongyi, founder of 360 Group, credits OpenClaw with completing a “nationwide popular science education,” making the public realize that AI agents are tools that can solve real problems. As BBC Chinese reported, the OpenClaw frenzy saw Baidu and Tencent hosting installation events with queues of nearly 1,000 people in Shenzhen alone.
Real-World Impact: From Steel Mills to Courtrooms
The commercial value of AI agents is now being validated across diverse sectors:
Industrial Manufacturing: Shougang Group has integrated AI agents into its steel production lines, slashing hot-rolling plan review time from one hour to seconds. Safety hazard identification agents have improved inspection efficiency by 60%. The Beijing Industrial Software Industry Innovation Center’s platform now boasts over 18,000 registered users, more than 1,200 AI agents, and over 5 billion industry data entries.
Software Development: CSDN founder Jiang Tao predicts that within one to two years, 99% of code will be generated by AI agents, with efficiency improvements of 3 to 5 times.
Content Production: Digital marketing professionals report that a single person can now produce over 10 advertising videos per day — a task that previously required a three-person team working three to five days.
Government Services: Beijing’s administrative reconsideration AI system has improved document generation and case review efficiency by 100%.
Zhao Xiangwei, director of the Informatization and Software Services Division at the Beijing Municipal Bureau of Economy and Information Technology, confirms that “AI agent applications are crossing the ‘from 0 to 1’ technology verification threshold and entering the ‘from 1 to N’ large-scale promotion stage. The returns on deployment have shifted from expectations to reality.”
The Cost Breakthrough
Perhaps the most critical driver of commercialization is cost reduction. Cao Peng, president of JD Cloud, offers a striking example: a business task that previously cost 2,700 yuan to run on a large model now costs just 10 yuan after inference optimization. “This cost advantage will become even more pronounced as token consumption increases,” he notes.
Consumer pricing has followed suit. Subscription models — weekly, monthly, or token-based — are emerging, with early adopters like Mr. Chen paying over 100 yuan per month. The model mirrors mobile phone plans, but with tokens replacing data allowances.
Policy Backing: A National Framework
The Chinese government has moved decisively to support this transition. The 2026 Government Work Report explicitly called for promoting next-generation intelligent terminals and agents. On May 8, the Cyberspace Administration of China (CAC), National Development and Reform Commission (NDRC), and Ministry of Industry and Information Technology (MIIT) jointly issued the first comprehensive national policy framework for AI agents.
According to an NDRC analysis, the policy releases three core orientations: balancing development with security, using application scenarios to drive technology, and promoting deep integration between agents and industry. It outlines 19 application scenarios across five domains — scientific research, industrial development, consumption, public welfare, and social governance.
Han Xia, executive deputy secretary-general of the MIIT Information and Communication Technology Committee, emphasizes the need to strengthen research on agent internet infrastructure, including system architecture, identity identification, and cross-platform collaboration protocols.
The Road Ahead: Challenges and Opportunities
Despite the momentum, significant hurdles remain. A technical report from the China Academy of Industrial Internet notes that while AI agents have moved from concept to a billion-yuan industry, challenges persist around standardization, security, and deep workflow integration.
McKinsey data cited in the research indicates that while 62% of enterprises are experimenting with agents, fewer than 10% have achieved large-scale deployment. Security concerns — including data leakage, unauthorized access, and malware distribution — have prompted Chinese authorities to issue formal risk warnings, particularly around open-source platforms like OpenClaw.
Chang Gaowei, initiator of the ANP Open-Source Technology Community, envisions the next phase: “AI group collaboration,” where agents work together autonomously across entire workflows. “The economic value it unleashes will far exceed what can currently be calculated,” he says.
Pan Helin, an MIIT expert, draws a parallel to the lithium battery electric vehicle boom: “In the next three years, agents will achieve geometric growth. The user base will explode, and the industry will enter a growth phase similar to what we saw with EVs.”
What to Watch
As China positions itself as the world’s most active AI agent market, several questions will shape the coming years: Will the regulatory framework accelerate or constrain innovation? Can costs decline enough for mass consumer adoption? And will the prediction of 99% AI-generated code materialize?
One thing is clear: the era of AI agents as a commercial reality, not just a technical curiosity, has arrived. For millions of Chinese workers and consumers, the question is no longer whether AI agents can work — but how quickly they will become indispensable.