Why one framework and not one per certification

The obvious way to build this would have been for each awarding body to define the competences its own credential assesses. It is faster, it needs no agreement with anyone, and it produces something that looks like a standard.

It also produces a market in which no two credentials can be compared. An employer holding two certificates has no way to establish whether they assert the same thing, and a learner moving between providers starts again. The word standard, in that arrangement, means only that a provider wrote something down.

The layer model

One framework defines the competences and the levels. Awarding bodies define credentials against it, publishing a coverage map showing which statements they teach and which they assess. Employers write roles against the same statements. Each layer does the work it is competent to do, and only the bottom layer needs to be shared.

That means this organisation does not award credentials, does not deliver training, and does not certify individuals. Those are the other layers, and they are open to anyone under a licence that permits commercial use without asking us.

What the reader should be suspicious of, given who wrote this

I founded this organisation, I hold an interest in a body that awards credentials against this framework, and I own a company that delivers training mapped to it. The argument above happens to describe a structure in which my other interests operate as licensees of a standard I published. That is a real conflict and it is declared above this article rather than at the foot of it.

The test I would apply, and invite others to apply, is whether the framework is written in a way that only one awarding body could satisfy. If a statement can only be assessed by an instrument that one body happens to own, that is capture, and the consultation process exists to catch it. Every statement is published with its indicators precisely so that the question can be asked from outside.

Why not simply defer to an existing standard

Where a competence is already well defined elsewhere, the framework should reference it rather than restate it, and the methodology says so. What did not exist was a level model for AI competence that credentials and job descriptions could both point at. That gap is the whole justification for this framework, and if another body closes it more credibly, the correct response is to converge rather than to compete.