The framework

Five dimensions, four levels. Behaviour-based and tool-neutral, so it does not go stale when the next model ships, and the same for everyone — what changes between a business owner, a public servant and a lawyer is the stakes and the examples, not the ladder.

Draft. A first draft, not yet reviewed. The framework, the questions and the wording are still being worked on.

The four levels

1. Orienting
Getting your bearings. You know this matters; you are not yet using it deliberately, and cannot yet say why it fails.
2. Practising
Using AI on real but low-stakes work, with growing confidence and the first habits of caution. Still learning when not to use it.
3. Capable
Using AI deliberately and selectively across regular work. Verifying as a matter of course, and making sound calls about when to reach for it.
4. Fluent
Sound judgment at speed — including the judgment of when not to use it at all. You take full responsibility for the result, and could help a colleague find their footing. Not tool-virtuosity.

The five dimensions, level by level

Understanding

What these tools are, what they can and cannot do, and why they sometimes produce confident nonsense.

Orienting: You know these tools matter and roughly what they do, but could not yet explain why one sometimes states a falsehood with complete confidence.
Practising: You can explain, in plain terms, that the tool predicts likely text rather than looking facts up — and that this is why it can be fluent and wrong at once.
Capable: You understand enough of how the models work to anticipate where they will struggle, and you adjust how you use them accordingly.
Fluent: You can explain the mechanism and its limits to a colleague, and you reason about a tool you have never used from how it works rather than from the marketing around it.

Working with AI

Choosing the right task to hand over, and directing the tool well — what to delegate and what to keep.

Orienting: You have tried a tool once or twice, or not at all, and are not yet sure what kinds of task it is actually good for.
Practising: You use AI for real but low-stakes work, and you are learning how to frame a request so the output is useful.
Capable: You choose deliberately what to hand over and what to keep, and you direct the tool well — giving it the context and constraints it needs.
Fluent: You match the task to the tool almost without thinking, know what to keep in your own hands, and get strong results without over-relying on the machine.

Judgment and verification

Telling good output from bad, checking what matters, and keeping your own thinking rather than anchoring on the machine.

Orienting: You tend to take the output at face value, and would not always know how to check whether it is right.
Practising: You have learned not to trust it blindly, and you look again when something feels off — though not yet as a routine.
Capable: You verify what matters as a habit, and you keep your own view rather than anchoring on the model’s first answer.
Fluent: Verification is reflexive and proportionate to the stakes: you can tell where the output is likely to be weak and check there first, and your own judgment stays in charge.

Responsibility

Confidentiality, privacy, the cross-border rule, the duties you cannot delegate to a tool, and owning the result.

Orienting: You have not yet thought through what is and is not safe to put into a tool, or who can see it once you do.
Practising: You know not to paste obviously sensitive material into a free chatbot, but the finer questions — privacy, the cross-border rule, what the provider may keep — are still hazy.
Capable: You think about confidentiality and privacy before you type, you know the duties you cannot hand to a tool, and you own the result.
Fluent: You handle sensitive information with a current grasp of the provider terms and the rules that bind you, and you treat the responsibility for any output as entirely your own.

Keeping current

Staying oriented as the technology and the rules move. A maintained state, never "done".

Orienting: You are not yet following how the tools and the rules are changing, and tend to hear about shifts after the fact.
Practising: You pick up the larger developments when they reach you, but not in any deliberate way.
Capable: You keep a deliberate eye on the changes that affect your work — a new model, a change to a provider’s terms, fresh guidance.
Fluent: You treat staying current as part of the job, not an extra, and you re-examine your own habits when the ground moves.

Drafted 21 June 2026 — a working draft, not yet reviewed. Guidance on AI capability, not advice on your own situation.