# One-Person Company
> VI: https://5ac.vn/wiki/one-person-company/ · EN: https://5ac.vn/en/wiki/one-person-company/ · lang: en · spec_as_of 2026-10-19 (spec no formal spec) · cluster C15
A business model where one founder runs the whole company through a role-based AI team on an agent operating system — commanding and approving instead of doing every job.
---
## Definition
A One-Person Company (OPC) is a business where a single founder runs every function — marketing, sales, operations, product development — with a standing AI team on an agent operating system. The founder's job shifts from executing tasks to issuing commands and approving decisions.
## Mechanism
The AI team is split into 14 business domains, and each domain carries three roles at three levels: a task-level agent that executes repeatable work, an analysis-level agent that measures and optimizes, and a strategy-level agent that proposes direction. The deployment 5ac.vn publishes runs 42 agent profiles across those 14 domains.
| Level | What the agent does | How the founder engages |
| --- | --- | --- |
| Task | Executes repeatable work to a brief | Hands off work in batches |
| Analysis | Measures, optimizes, summarizes reports | Reads the dashboard daily |
| Strategy | Proposes direction backed by analysis | Approves the decision |
The operating principle is "command, not execute": set the objective, assign the right agent, review the output, then refine and repeat. A founder does not need to write code to ship a feature, or to keep the books to know where cash stands.
Work runs asynchronously: the founder hands off and reviews inside designated decision windows while agents execute between them, with no real-time meetings required. The coordination layer — Kanban, cron, a knowledge base, and a KPI dashboard — turns scattered outputs into work with an owner, a status, a log, and an approval point.
A seven-metric KPI board, from agent-influenced revenue and lead response time to on-time task rate, net hours saved, and the founder's own load, shows which parts of the system actually produce return. The biggest risk is losing command, so permissions follow least privilege: an agent touches only the data its task needs, high-risk zones stop at drafts, and every action is logged.
Flow: Founder commands → Agents execute → Review and approval → KPI dashboard
## Application context
This is a term small-business readers ask about directly — "what is an OPC?" and "can one person really run a company?" For Vietnamese founders, 5ac.vn maintains a dedicated pillar on the model plus a companion KPI piece, and G-Company OS is the operating system the company built for exactly this model. The usual rollout is a 30-day playbook: week one lists the most repeatable tasks, agents get assigned to high-frequency low-risk work, work moves into a queue, and return is measured in week four. The standing rule: measure before you scale, cap permissions before you automate.
## Sources fetched for this entry
- — OPC pillar: 14-domain architecture and the command principle — accessed 19 Oct 2026
- — the seven KPIs for a solo founder running an AI team — accessed 19 Oct 2026
- — 30-day playbook and the loss-of-command risk — accessed 19 Oct 2026
- — English OPC pillar — accessed 19 Oct 2026
- — English KPI dashboard piece — accessed 19 Oct 2026
- — English playbook piece — accessed 19 Oct 2026