Prompts and "Reverse Engineering"—The Idea of Designing from the Desired Answer (Generative AI Prompt Textbook 8)
The more accustomed people become to using AI, the more they hit this wall.
"Close, but not quite right." That feeling of "groping in the dark" as you tweak and test a prompt dozens of times. Generative AI is certainly excellent, but there is no such thing as a "perfect one-shot answer" no matter how far you go.
But it is time to realize something. There is an approach in the opposite direction: instead of seeking an answer, you design the question by working backward from the answer.
When thinking about how to write a prompt, many people focus on "How should I ask to get a good answer?" That is natural. But if you dig one level deeper, you arrive at the perspective of "What kind of answer is 'desirable'?"
For example, suppose you want AI to create a project proposal for you. But what you really want isn't the proposal itself. It might be "a persuasive plan that will get approved by my boss on the first try," or "an external perspective that helps organize my own ideas."
In other words, what you really want is not the "deliverable," but the "state that the deliverable brings about."
Reverse-engineering prompts are a design method that works backward from that "goal emotion," "expected effect," and "context of use."
✅ Basic structure:
1. In what situation will it be used?
2. Who will be reading it?
3. How do you want the reader to feel?
4. To achieve that, what kind of structure/wording is necessary?
If you ask yourself these questions first and then give instructions to the AI, you will get a completely different output.
✅ Reverse-engineering prompt example (presentation script)
Please create a 3-minute speech script for a pitch event tomorrow. The audience is investors. They don't have much time, so I need to make them think "this sounds interesting" in the first 30 seconds. Focus more on the scale of the problem being solved and the size of the market rather than technical details. Finally, include an impressive phrase that they will remember in one go.
—Like this, you decide the usage situation, the audience's reaction, and the key points of the structure first, then design the question. This is the prompt technique of reverse-engineering thinking.
Reverse engineering has another important meaning. That is "clarifying your own purpose."
Vague questions lead to vague answers.
Conversely, if you are clear about "what you want to do," "what you want to convey," and "what situation you will use it in," the AI will "read" that intent.
This is finally the moment when the AI changes from a "slave to words" into an "assistant for thinking."
✅ Reverse-engineering prompt example (SNS post)
Please think of a post for social media today. The theme is "How to bounce back after failing at work." The followers are mainly working women in their 20s and 30s. I want to attract empathy, so start with a description of a failure story, and make sure it leaves a slightly positive lingering feeling at the end. Also, add 3 hashtags.
—This, too, is reverse engineering from "how it will be used" rather than "what to put out."
The important thing is that a prompt is not a "question," but can be a "blueprint for output."
Since it is a blueprint, it is perfectly fine to start drawing from the finished image.
Until now, we have tried to get answers by "asking" the AI. But from now on, the ability to envision the "answer you want to give" first and design the question to draw it out will become the writing technique of the AI era.
That is no longer just "prompt technology." It is the design of thinking and the construction of communication itself.
Next time, the theme will be "Prompt 'Temperature'—How to Infuse AI with Passion and Tension?" You cannot command emotions. However, you can create an atmosphere. I will reveal the secrets of prompts that imbue words with temperature.