Zero-Person Company

Definition

A zero-person company is an experimental operating model in which agents perform most recurring research, production, publishing, distribution, observation, and iteration, while a human remains outside the routine execution loop as goal setter, system designer, risk owner, and final authority.

Current Synthesis

“Zero-person” describes the execution layer, not the absence of human agency. The source’s media experiment automates research, writing, publication, traffic attribution, analytics, and feedback, but depends on a human’s prior writing, conversations, comparative habits, standards, and decisions about what the system should optimize. The system can continue operating without step-by-step supervision only because those judgments have been externalized into Context Infrastructure, prompts, files, metrics, and AI Skills.

The commercial claim remains unproven in the source. The project generated audience growth and contributed indirectly to training revenue, but had not independently discovered and closed a new business loop. This distinguishes Zero-Person Company from a literal autonomous firm and from One-Person Company: AI can compress execution and coordination, yet demand discovery, responsibility, trust, pricing, and customer delivery remain unresolved human and institutional functions.

Key Claims

• The strongest current form is unattended routine execution under human-defined goals and constraints, not a company with no human responsibility.
• Agents can automate connected loops across research, production, publishing, attribution, analytics, and revision.
• Durable autonomy depends on externalized context, standards, operating procedures, metrics, permissions, and escalation rules.
• Audience or workflow automation is not equivalent to a closed commercial loop with independent demand, payment, and delivery.
• Existing domain, commercial, editorial, or operational judgment determines how much leverage the system can create.
• Risky external actions still need authority boundaries, auditability, and human review proportional to failure consequence.

Evidence

• Automated media loop: “有了AI，我感觉自己强得可怕！”｜对谈鸭哥 describes AI handling collection, research, writing, publication, channel attribution, analytics, and iterative adjustment for a newsletter and social account.
• Human judgment substrate: “有了AI，我感觉自己强得可怕！”｜对谈鸭哥 says differentiated output depends on the operator’s articles, chats, worldview, comparison methods, goals, and acceptance standards.
• Commercial boundary: “有了AI，我感觉自己强得可怕！”｜对谈鸭哥 reports indirect training income but says the project had not independently completed a new business loop.
• Governance boundary: “有了AI，我感觉自己强得可怕！”｜对谈鸭哥 retains human review for consequential publication and rejects high-risk physical automation where one error could be unacceptable.

Counterevidence & Qualifications

The term can obscure ongoing human labor in system design, context curation, quality judgment, exception handling, legal responsibility, and commercial strategy. The source supplies one operator’s experimental project rather than evidence that autonomous firms can reliably acquire customers, contract, pay taxes, maintain trust, or bear liability. Reported audience growth also does not establish durable profitability or transferability to other operators.

What Changed

• Added a bounded definition that separates routine execution autonomy from legal, commercial, and moral autonomy.
• Captured the difference between audience automation and a proven end-to-end business loop.
• Made externalized judgment, escalation, and risk ownership explicit prerequisites.

Related Concepts

• One-Person Company - neighboring model in which one human remains the visible operator across the business loop.
• Context Infrastructure - substrate that externalizes goals, knowledge, standards, and feedback for unattended execution.
• Agentic Workflow - action-observation-revision loop used to carry out recurring work.
• AI Organization Design - broader allocation of goals, roles, review, and accountability across humans and agents.
• Agent Permission Boundaries - limits on actions whose consequences exceed the system’s authority.
• Output Quality Gates - criteria that determine whether generated work can progress or be published.
• Human Judgment Under AI - continuing responsibility for demand, meaning, trust, and acceptance.