AI Reshapes Corporate Structures, Hollows Out Management

Artificial intelligence is fundamentally restructuring corporate hierarchies worldwide, hollowing out middle management layers while demand for high-end specialized talent remains stable, according to a landmark report jointly released by Chinese recruitment platform Tongdao Liepin and Tsinghua University’s School of Economics and Management.

The “Talent AI: Skills Trend Report in the AI Era” (Caixin Global), launched at Tsinghua University on June 17, documents a phenomenon termed “middle management collapse” (中层塌陷), where AI-driven automation is systematically eliminating mid-level white-collar positions while companies report record profits.

The Middle Management Crisis

Dai Kebin (戴科彬), CEO of Tongdao Liepin Group, told the conference that AI has entered what he called a “daily update era” in 2026, systematically reshaping employment structures and organizational forms globally. “Under AI and automation, mid-level white-collar positions are under significant pressure, while high-end specialized and leadership talent remain relatively stable,” Dai said, as reported by Dianshangpai.

The data is stark. In China alone, demand for entry-level sales roles fell 14.47% in the first five months of 2026, business roles dropped 10.50%, and brand roles declined 9.74%. On Liepin’s platform, machine learning jobs requiring less than one year of experience plummeted 71.43%, while image algorithm roles fell 66.67%.

The Global Layoff Wave

According to TechCrunch, 363 tech layoff events have occurred in 2026 affecting approximately 150,000 workers. AI has been the most-cited reason for layoffs for three consecutive months, with an average of 974 people laid off per day — 44% faster than the previous year.

What makes this wave distinctive is what analysts are calling the “profit paradox.” Companies cutting jobs due to AI are simultaneously reporting record profits. Amazon’s net profit surged 77% while cutting 30,000 jobs, 78% of which were middle managers at levels L5-L7. Oracle cut 21,000 jobs while quarterly net income rose 27% to $3.7 billion. Cloudflare eliminated 1,100 positions — 20% of its workforce — while reporting its highest single-quarter revenue ever at $639.8 million, up 34% year-over-year.

Cloudflare CEO Matthew Prince explicitly stated that the “vast majority” of laid-off employees were “measurers” — middle management, finance, legal, and internal auditing roles — whose functions had been automated by AI systems.

The 7×4 Skills Framework

The report introduces a novel 7×4 analytical framework comprising seven AI skill domains and four competency levels, built through a dual deductive-inductive process referencing international taxonomies and Chinese enterprise data. The seven domains are: AI Foundation Algorithms & Models, Generative AI Applications, AI Agent Construction, AI Multimodal Understanding & Generation, Physical AI, AI Data/Computing/Engineering Deployment, and AI Ethics, Safety & Compliance.

Professor Xu Xin (徐心) of Tsinghua University noted that the framework reflects two underlying logics: technology evolution from perception to generative AI to agents to physical AI, and the parallel imperative of development and governance. The inclusion of “AI Data/Computing/Engineering Deployment” as a standalone domain is a unique contribution of this research, as international reports have not separately identified it.

From People Cost to Token Cost

Dai Kebin outlined five levels of AI’s impact on organizations, from individuals managing multiple AI agents simultaneously to the breakdown of departmental silos. Perhaps most strikingly, he predicted that HR departments will soon calculate “Token Cost” rather than “People Cost” — how many GPT accounts a team uses and how many tokens are consumed are becoming new metrics for organizational efficiency.

“The most important change in human history,” Dai said, as reported by ifeng.com, “is that skills are beginning to separate from people, breaking the paradigm of human skill generation and inheritance. Organizations can have people come and go, but skills are permanently precipitated.”

China’s Distinct Path

The report highlights China’s unique position on Physical AI, which ranks fourth among the seven skill domains in China, with demand three times that of AI agent construction in major cities like Beijing, Shanghai, Shenzhen, and Hangzhou. This reflects China’s long-standing “digital-real integration” national strategy and distinguishes Chinese AI development from international patterns.

The Guardian Model

Professor Guo Xunhua (郭迅华) of Tsinghua University offered a compelling framework for understanding the evolving human-AI relationship in the workplace. He described the relationship between humans and AI agents as one of “guardianship.” “Agents have ability and can learn but cannot take responsibility — like a gifted teenager,” Guo said. “Humans must do three things: set goals, nurture and gatekeep, and take responsibility.”

What’s Next

As AI continues to reshape corporate structures, the report points to both challenges and opportunities. While AI eliminates some roles, embodied intelligence is creating new long-chain industries, with emerging positions including Agent Product Manager, Prompt Strategy Designer, AI Governance Expert, and AI Workflow Designer on the white-collar side, and Robot Inspection Technician, Human-Robot Collaboration Safety Supervisor, and Industrial AI Training Data Collector on the blue-collar side.

Professor Yang Bin (杨斌) of Tsinghua University offered a cautionary note, introducing the concept of “AI to the power of n” (AIⁿ). He warned that if an organization’s “base number” — its fundamental structure and capabilities — is less than one, applying AI as an exponent will only amplify its weaknesses. The real challenge for enterprises, he suggested, is not how powerful AI is, but whether they can complete their own reconstruction under AI’s influence.