The zero-person company.
A business where the org chart is an agent board. Where decisions are drafted by models, approved by humans, and executed by software. Where leverage per founder is 100× what it was in 2022. This is not science fiction. The components exist. Most founders just haven't assembled them yet.
The pieces: context layer · agent executives · human approval · execution rails.
Four pillars make it possible.
Each one is a solvable engineering problem. We build all four.
Context layer
The system of record for what the business knows. Documents, decisions, metrics, and frameworks — legible to agents. Without this, every agent is generic.
Agent executives
Role-scoped agents — CMO, CFO, COO, Chief of Staff — trained on the business, operating within clear guardrails. Not fully autonomous. Always human-approved on material actions.
Execution rails
Agents that can act need safe tools: tooling APIs, deployment pipelines, CRM write paths, payments, email, calendars. The rails decide what they can touch.
Human approval fabric
Every material action flows through a human. Auditable logs, accountable owners, clear escalation. This is the fabric that makes the company safe, not optional.
Where to start.
The wrong move is an "AI everything" program. The right move is one workflow, or one role. Prove the pattern. Expand from what worked.
Start with one workflow — pick one high-value workflow (deal briefs, proposals, renewal risk). Ship it end-to-end in 4–6 weeks.
Start with one role — pick one exec function (CMO, CFO, Chief of Staff). Build an agent that operates it — with human approval — in 4–6 weeks.
Start with a maturity baseline — AEMI: a 30-day assessment of where your AI adoption actually stands, with a prioritized roadmap to a mature context layer.
Questions we get on this
Is this a real product I can buy today?
No. It's a thesis about where this is heading, and a set of capabilities MetaCTO is building toward. Today we ship the components — AEMI, Enterprise Context Engineering, AI Exec Team — that together move a company in this direction.
Does MetaCTO think humans will be removed?
No. We think the ratio will change. Ten humans doing the work of a hundred. A hundred humans running what used to require a thousand. The role of humans stays — judgement, relationships, ethical accountability. Execution and analysis shift to agents.
What has to be true for a zero-person company to actually work?
Four things: a production-grade context layer so agents can reason over the business, role-scoped agents with clear guardrails, a human-in-the-loop approval fabric, and execution rails (tools, APIs, deployment pipelines) that agents can safely use. We build all four.
Is this safe?
Only with human approval gates on material actions, auditable logs of every agent decision, and clear accountability mapping to real humans. We do not build systems that act unilaterally in consequential domains. This is engineering, not utopianism.
What is the right place to start?
Not with an "AI everything" program. Start with one workflow (via Enterprise Context Engineering) or one role (via AI Exec Team). Prove the pattern. Then expand. Most programs that try to start with "the whole company" stall.
Who is writing about this?
Sahil Lavingia's 'one-person company', Aaron Levie and Marc Benioff on agentic enterprises, Shopify's agent-first mandate, various essays in Every and Stratechery. We think the shape is right. The timing is imminent.
Who this conversation is for
Good fit
- Operator or founder thinking seriously about AI-native org design
- Already running AI workflows and ready to plan the next 2–3 years
- Interested in building the context layer now, not retrofitting it later
- Want a working conversation, not a pitch deck
Not a fit
- Looking for a framework to ban AI from the company
- Expecting to buy a 'zero-person company' off the shelf
- Need immediate ROI from one project (start with Agentic Workflows)
- Not ready to engage before the landscape settles
Let's talk about the thesis.
Bring questions, pushback, a real use case. We will tell you honestly what is possible today and what is two years out.