An old Catherine Tate sketch offers a surprisingly useful way to understand the AI skills paradox.
In the sketch, a woman claims she can interpret seven languages.
To someone who does not speak those languages, her confidence and delivery might resemble expertise. To the people who do, she is making meaningless noises.
AI creates a modern version of this problem.
It can make someone sound more capable before it makes them more capable.
Fluency can look like expertise
AI can give almost anyone the vocabulary, structure, polish, and confidence associated with expert work.
Sometimes that creates genuinely useful new capability. A person can learn faster, explore unfamiliar ideas, improve a draft, or complete work that would previously have been out of reach.
But fluent presentation is not the same as sound judgement.
A polished answer may still contain weak assumptions, missing context, invented evidence, or conclusions that do not fit the situation. The problem is that those weaknesses can become harder to see when the work sounds confident and complete.
That matters because people without subject-matter knowledge may be unable to distinguish an excellent AI-assisted answer from a fluent but fundamentally wrong one.
The people who already understand the work can usually see the difference.
They speak the language.
AI can scaffold, accelerate, or camouflage
The value of AI depends partly on the capability and self-awareness of the person using it.
- For a learner, AI can be a scaffold.
- For an experienced practitioner, it can be a powerful accelerator.
- For someone who lacks both expertise and self-awareness, it can become camouflage.
Those are very different outcomes from the same technology.
A learner may use AI to ask questions, test understanding, and build capability. An experienced practitioner may use it to move faster while applying judgement that comes from years of work. Someone without enough knowledge may accept the output because it looks convincing.
The tool can support all three people.
It cannot guarantee that they know which result is good.
Expertise moves into judgement
This is why AI does not eliminate the need for expertise.
It changes where expertise becomes most valuable.
Producing a first draft becomes easier. Judging whether it is correct, relevant, safe, and useful becomes more important.
Experts recognise what is missing. They notice the assumption that should have been tested, the evidence that does not support the conclusion, the edge case that changes the answer, or the operational constraint the model cannot see.
That judgement may be less visible than producing the original work, but it becomes more important when polished output is cheap and abundant.
A practical AI literacy check
AI literacy is not only knowing how to write prompts.
It includes knowing:
- what good work looks like
- which assumptions need testing
- what evidence is missing
- where the likely failure points are
- when a qualified expert must be involved
Organisations need that capability around the AI system, not only inside the prompt.
That may mean stronger review points, clearer evidence requirements, named subject-matter owners, or explicit escalation when the work moves beyond the user's competence.
Without those controls, an organisation may simply produce confident mistakes faster.
With them, AI's speed and human judgement can create better work at a scale neither could achieve alone.
The most important question is no longer only:
“Can AI perform this task?”
It is:
“Who in the system can recognise when it hasn't?”