The Working Parts of an AI-Native Organization

An AI Operating Core (AIOC) has a versioned Blueprint, a human-facing Panel and a Back office. The Blueprint describes the work, input and output contracts, components and fixed tests. The Blueprint is one versioned file from which the Panel and the Back office are built. The Panel is the unit’s own control screen, written in code, from which every task of the unit can be controlled. The Back office is the AI chat workspace where the unit’s AI works; it keeps the Panel, also after the Panel is built, and can rebuild it if it fails. A stop, a signature or a report to a human never depends on the Back office.

A Cell has a stable identifier and defined responsibility. It is filled by AI or held as a human position, and what fills it can change. Changing the filler does not by itself change the role’s contract, acceptance criteria or authorization, and a move between AI and a human position must still respect the Human Floor. A human position is a Cell of human nature: it is defined only as human, so that no AI fills it, and by its duties. Who holds it is decided by the organization’s own mechanism for its human side, outside the system. This lets capability evolve without erasing accountability.

A Pod gathers Cells around a function and presents a coherent outer interface. Its member Cells retain their own access and responsibilities. A Pod groups Cells only and never holds another Pod. Anything that needs nesting is an AIOC, which can hold other AIOCs to any depth and still looks like one piece from outside.

The core is the first unit of an organization’s AION. It holds the governance seat, the organization’s own rules and a governance panel. The tool gateway and the stop are installed with the core but run as code beside it, outside the reach of every AI, the core’s own included. Under it, Top Management builds and runs the operation from the approved design; the AION designs and proposes, and never runs the operation. Top Management Group, the first unit under the core, is the hub of the operation, for people and for AI. Every human position has its own Chief of Staff, which leads the unit’s AI agents and brings the position only what needs a decision. It may order and summarize what reaches the position but never drops or delays any of it, and the raw record stays open to the position. Every call AI units make to a tool passes through the tool gateway.

Typed ports specify what a part may receive or return. A Work Item states the requested result, sources, owner and acceptance test. The receiving part checks its incoming work; a produced answer does not count as complete until the relevant acceptance conditions are met.

Imagine a report-producing Cell replaced by another authorized performer. Its fixed role and test remain legible. If the requested change affects access, scope or protected decisions, the appropriate human position must authorize it. The Blueprint records the new version.