Before You Prompt

Artificial intelligence is very good at turning uncertainty into structure. You arrive with a vague problem, a half-formed idea, a paragraph that does not work, or a decision that feels too large, and the system immediately gives you something shaped: a list, a draft, a summary, a plan, a tone, a direction.

That is part of its power. It is also where the danger begins.

The problem is not that AI thinks for us in some simple replacement sense. The problem is that it often starts organizing the problem before we have understood what kind of problem we are bringing to it. It gives form to ambiguity. It names categories. It proposes priorities. It makes the next move appear obvious.

This can be useful when the human has already established a position. It can be distorting when the machine becomes the first frame.

Before you prompt, pause long enough to know what you are actually asking for.

This does not need to become a ritual, a system, or another layer of productivity theater. It can be very simple. Open a blank note and write three things before involving AI: what you know, what you want, and what must not happen.

What you know is the ground. It includes facts, constraints, context, doubts, deadlines, names, numbers, political sensitivities, business limits, aesthetic preferences, or anything else that belongs to the situation before the machine enters it. This matters because AI is often strongest when the user has already defined the territory. Without that territory, the model will invent a plausible one.

What you want is the intention. Many weak AI interactions begin with a vague instruction because the user has not decided whether they need a draft, a diagnosis, a counterargument, a structure, a decision, or simply a better question. “Help me with this” is sometimes enough, but it often invites the system to choose the mode of work for you. That choice should remain human.

What must not happen is the boundary. This may be the most important part. Do not make it generic. Do not make it sound like a corporate memo. Do not flatten the argument. Do not remove the tension. Do not over-explain. Do not turn this into marketing. Do not pretend certainty where there is only a hypothesis. These limits protect the work from the machine’s tendency toward acceptable completion.

Only then prompt.

The difference may seem small, but it changes the relationship. AI stops being the origin of the frame and becomes a force applied to a frame you have already built. It can expand, compress, challenge, translate, simulate, or sharpen. It can help you see alternatives. But it is no longer the first author of the situation.

This is especially important in writing. When AI begins too early, it can produce a competent version of a thought that has not yet matured. The result may be fluent, balanced, and usable, but strangely empty. It may resolve the tension that should have remained alive. It may smooth out the friction that contained the real idea.

The same applies to strategy and decision-making. AI can make a weak idea operational. It can give a professional surface to an unclear priority. It can produce a roadmap before the destination has been honestly chosen. In that sense, the machine does not only accelerate good thinking. It can also accelerate premature thinking.

The discipline is not to avoid AI. The discipline is to enter the exchange with enough human orientation that the system has something real to work against.

Before you prompt, ask yourself: what do I already believe about this? What am I not sure of? What kind of answer would be dangerous because it would be too easy to accept? What should remain unresolved for now?

The better future of AI work may not depend only on better prompts. It may depend on better pre-prompts: the private, human moment before the visible instruction. The small interval in which we decide what the machine is allowed to shape, and what must remain ours.