Four months until EU AI Act enforcement. The documentation does not exist.

The EU AI Act isn't a future concern you can plan around later — high-risk system providers are already expected to maintain technical documentation covering system architecture, data lineage, human oversight mechanisms, and performance characteristics.

Most organisations cannot produce this documentation. Not because they haven't tried, but because the tooling to create it as a first-class output of the design process has never existed.

What most teams have instead is architecture diagrams in Lucidchart, governance notes in Confluence, and compliance checklists in a spreadsheet — three places that don't reference each other, and none of it holds up under audit.

AI systems have gone into production across most enterprises over the last three years — models integrated, agents built, pipelines automated — usually faster than legal, risk, and compliance teams could keep pace.

The result is a documentation debt that is now coming due. Organisations are being asked by regulators, auditors, and boards to explain AI systems that were built without governance as a design consideration.

Retroactive documentation is expensive and often wrong by the time it's finished — the people who built the system have moved on, and the system itself hasn't stood still.

Enterprise architects already know the problem and have tried to solve it with whiteboards, diagramming tools, and documentation templates — none of which produce governance output, carry typed semantics, or answer the question regulators are actually asking.

That gap turns every AI governance exercise into a manual, one-off project, disconnected from the actual architecture it's supposed to describe.

ModelAIr is built to close that gap — a design workbench where governance is the direct output of drawing the system, not an extra layer stapled on afterward.