Prompting Is Daring to Give Orders
Watch someone who is genuinely good at prompting an AI, and you will notice something that has little to do with technical skill: they are comfortable giving orders. They state what they want, plainly, to a system that will not judge them. The people who struggle with AI are rarely struggling with the interface. They are struggling with the act of command itself — the same hesitation that makes it hard to delegate, to direct, to say clearly what should be done.
I saw this vividly in two conversations around the same time. One was with a ride-hailing driver who wanted his son to study IT because “people in IT are smart.” I mentioned that AI can now write code on request — and his reaction was not curiosity but something closer to fear. The other was at work, where I had quietly started using AI for most of my tasks on an internal project with no confidential data. When my manager found out, the response was anxiety about security, followed by the instruction that I should have asked permission first.
The contrast was instructive. A senior colleague — far more experienced than me, from a background building mobile apps at a major bank — was struggling through the same project coding raw, in a bare editor with no extensions, copying snippets into a chatbot one paste at a time. He is not unintelligent; he is afraid. Afraid of violating policy, afraid of the new workflow, afraid of being seen relying on a machine. Meanwhile the work itself was CRUD screens and forms — the kind of thing that could honestly be solved with a spreadsheet. Four people were allocated to it. One person with real AI tooling could have finished it in a fraction of the time.
The pattern generalizes beyond any single office. Security theater is the respectable excuse, but underneath it sits something rarer and more decisive: agency. Most people do not dare to tell a machine — or anyone — what to do. They wait for permission, for process, for someone else to move first. And permission-asking culture compounds: managers who micromanage, review every line by hand, and split work into fragments produce teams that are slower than any individual member could be.
There is a fair objection here. Daring to give orders is only a virtue when paired with judgment about which rules actually matter. Ignoring security policy on genuinely sensitive systems is not boldness; it is negligence. The skill is telling the difference — knowing which constraints protect something real and which exist mainly so that no one can be blamed later.
But the deeper point stands, and it explains why AI adoption is splitting people into two groups faster than any previous technology. The tools are available to everyone. What is scarce is the willingness to command: to describe an outcome, accept responsibility for directing the work, and let go of the fear of doing it wrong. Prompting looks like a typing skill. It is actually a posture toward authority — including your own.
Drafted with AI assistance from my personal journal (2025-09-17), then edited by me. See On AI Assistance.