Registry · ActiveFiling /MAI-001Issued 26.iv.2026
§ 01The matter at hand

Subject

The documentation gap.

Enterprise AI systems operating without governance record.

Aug 2026 — four months

EU AI Act enforcement.

Nobody knows what their AI is allowed to do.

An architecture design workbench in which the diagram is the interface, the schema is the governance, and the artefact is the record.

§ 02The argument

Position

Governance is not a layer bolted on. It is a property of the design.

The diagram is the mechanism. Governance is the product.

Every architecture team has diagrams. They live in Lucidchart, Miro, draw.io. They get exported to PDF, attached to compliance documents, and quietly frozen in time. The systems they describe keep moving. The diagram does not.

A regulator does not want a picture. A regulator wants a structured record. Which models are deployed where. What authority each component holds. Where human oversight is required. What classification of data flows through which connection. These are properties of the system, not properties of a drawing of the system.

The fix is not better diagrams. The fix is to treat the design itself as a structured record — one where placing a component on the canvas declares not just its shape but its authority, the data it touches, and the conditions under which a human must be in the loop. The diagram and the documentation become the same artefact, because they were never meant to be different.

In the field

Most enterprises hold AI architecture in three places: a diagramming tool, a Confluence page, and a risk register. None of the three reference the others.

What this is not

Not Lucidchart. Not Visio. Not LeanIX. Those produce documents. This produces a record.

§ 03The mechanism
/01Typed node schemaComponents as structured objects.Every component is a typed object with domain-specific properties. Not a shape. A record.
/02Assertion levelsSignal. Advisory. Authoritative.Three levels define what a component is governed to do — not what it can do. The distinction matters legally.
/03Authority boundariesWhere autonomous authority ends.Explicit modelling of where AI acts alone, and where a human must approve. Visible in the diagram. Exportable.
/04Edge semanticsConnections that carry meaning.Transport, authentication, sensitivity classification, reliability — every edge is a typed governance object.
/05Governance outputThe diagram is the artefact.No export step. No documentation layer. The design workspace is the governance record.

Status

Beta · Invite only

Capacity limited


For architects under governance pressure.

Built for architects, not curious observers.

If you are under pressure to produce AI system documentation that does not yet exist — whether for the EU AI Act, an internal audit, or a regulator who has stopped accepting promises — we want to hear from you.

Request accessNo commitment · Early access only