SearchFree Newsletter

Atomic Scaling framework

HAAO — the Human-Assisted Autonomous Organization

Definition

A Human-Assisted Autonomous Organization (HAAO) is a company in which AI agents are the default operators and humans govern the exceptions — a term coined by Ludovic Bodin of Atomic Scaling to describe the org chart that replaces the human-run company.

The problem it names: Teams bolt AI onto a company designed for humans, and get demos instead of operating leverage.

In a conventional company, humans run everything and use tools. In a DAO, humans govern collectively through code. In a HAAO, autonomous agents are the default operators and humans step in only when needed. The org chart still exists — it is upside down.

Three communication layers

A HAAO runs on three channels. Most companies have only ever designed the first.

H2H

Human to Human

The layer that already existed. In a HAAO it shrinks — fewer status meetings, because status is a query against a system that is always running.

H2A

Human to Agent

The new managerial craft. Not prompting: specification, context curation, and the design of meaningful interrupts. The 2023 prompt engineer is the 2026 agent designer.

A2A

Agent to Agent

Agents reviewing and dispatching to each other on a shared surface humans can read and interject into. This is where throughput comes from — and where unsupervised drift starts.

What stays human

Humans in a HAAO are not removed. They are repositioned from operators to governors, and what they hold is JTO — Judgment, Taste and Override: the ambiguous call, the quality bar, and the authority to stop an agent that is confidently wrong. An organisation that automates without naming those three roles has not built a HAAO. It has built an unsupervised one.

The honest version of this includes the failure: run agents unsupervised for two nights and the second night they mix projects, code and architecture. It took us the better part of a week to recover. That week is the argument for override.

Next step

How ready is your company to run on agents?

The Atomic Scaling Score reads your company across the six pillars of the 3P3R Method® — People, Prediction, Playbook, Reach, Retention and Revenue — and names the constraint to fix first. Free, two minutes.

Take the Atomic Scaling Score →See the Atomic Scaling OS

Questions people ask

What is a Human-Assisted Autonomous Organization?

A Human-Assisted Autonomous Organization (HAAO) is a company in which AI agents are the default operators and humans govern the exceptions. The term was coined by Ludovic Bodin of Atomic Scaling. It inverts the traditional org chart: agents at the base running continuously, humans at the top holding judgment, taste and override.

What does HAAO stand for?

HAAO stands for Human-Assisted Autonomous Organization.

How is a HAAO different from a DAO?

A DAO — Decentralized Autonomous Organization — distributes human governance through code and token holders. A HAAO is not about decentralisation: agents do the operating work, and a small number of named humans govern the exceptions. One redistributes human decision-making; the other reduces how much of it the business needs.

How is a HAAO different from a One-Person Company?

A One-Person Company (OPC) is a single human on top of an agent stack — full autonomy, but a hard ceiling on complexity and an extremely draining one to run. A HAAO pairs a small team with agents, which removes that ceiling. Atomic Scaling treats HAAO as the higher-order play.

How many AI agents can one person manage?

In practice Atomic Scaling sees each human managing at least two agents, and experienced operators far more. The limit is not tooling — it is whether escalation triggers and override authority have been defined. Encode those before scaling past roughly ten agents per human, not after.

What is the difference between prompting and looping?

"Stop prompting, start looping" is the Atomic Scaling shorthand for the shift from asking an agent for an output to running a continuous system with work always in flight. A prompt produces an answer; a loop produces an operating capability that keeps running when you stop typing.