In the early days of generative AI, a lot of prompting advice focused on assigning roles, defining constraints, and specifying output formats. "Act as an editor." "Act as a strategist." "Use this structure." "Return the answer in this format."
That made sense. Older models often needed a carefully drawn magic circle. Without enough framing, they might answer the literal question while missing the transformation the user actually needed.
If Cinderella said, "I want to go to the ball, but I have no carriage," an older AI might have explained transportation options.
So the user had to write the whole spell: You are a fairy godmother. Cinderella wants to go to the ball. Available materials include a pumpkin, mice, and lizards. Transform these into transportation. Use the following format.
That was not just prompting. It was system design.
The center of gravity has moved
Newer conversational AI changes where that work happens. The user does not always arrive with a perfect prompt. Often, they arrive with fragments: frustration, intuition, half-formed ideas, constraints, rejected directions, and a sense that they need to move somewhere else.
A useful AI system can listen across that conversation, connect scattered dots, infer the direction of travel, and become useful for the next step.
That is why "just talking to AI" can be more effective than writing a perfect prompt.
Fluency, in that light, doesn't always look like a good prompt. Sometimes it looks like barely a prompt at all — just thinking out loud, in the room, with something listening closely enough to catch it.
The wish, not the blueprint
The magic is not in following a fully specified wish. It's in helping the user find the wish inside the fragments — and turning those fragments into the next move.
This may be one way to think about AI fluency. Not the ability to write better instructions — the ability to bring your unfinished thinking into the room, and let the shape of the task emerge as you go.
We used to write magic blueprints. Now we can start with a wish.