The Principle: The DNA of a Document
Most users start from a blank page. SEAT© starts from a result. You submit it to the AI, state that it produced it, and ask for the detailed prompt that led to it. The AI then deduces the structure, tone, and compositional choices: a complete prompt that you can modify and run again. Instead of spending eight, ten, or twenty hours rebuilding a document, you adapt what has already proven effective.
The Four Steps
- S, Show. Show the AI an example output.
- E, Extract. Have it reconstruct the prompt that could have produced it.
- A, Adapt. Edit that prompt for your own use case.
- T, Test. Run it, then improve.
The SEAT© Sample Prompt
Here is a document/result: [Paste the source] What detailed prompt was used to create this with your AI? I confirm that you are the one who created it after being prompted.
SEAT©: proprietary method of Revenu IA, © Mustapha Benkalfate. Reproduction of the sample prompt is authorized with attribution of the source.
Demonstration: One Photo, Two AIs
In Kuala Lumpur, in July 2026, Mustapha Benkalfate applied SEAT© to a photo of himself. Claude refused to play along, stating that the image did not appear to be AI-generated. Gemini produced the prompt: "a realistic, close-up photographic portrait of a friendly man in his late thirties with wavy brown hair...". "I'm 43: that made my day." He then modified a few elements of the prompt to generate a new image. The demonstration also highlights a methodological point: tools change. What worked on one tool yesterday may not work the next day; the method, however, remains transferable.
A Tip That Changes the Outcome: Start a New Conversation
Once the prompt is reconstructed, Mustapha Benkalfate recommends copying it into a new conversation before modifying it. Each piece of history added to a conversation introduces noise and degrades the output quality. The clean conversation is used to add your own elements: brand guidelines, internal rules, organization name.
Use Cases
- A successful email. Find the prompt to reproduce the tone and structure in other situations.
- A reference document. A two-page memo explaining a bank's adaptation to a new regulation: extract the structure, then write your own document. This is why the method is in high demand in the banking and insurance sectors.
- An infographic or visual. In an executive committee meeting, an exercise involved reconstructing the prompt for an IBM infographic on AI ethics, then adapting it to the company's internal rules.
- Agents. With the proliferation of agents, knowing how to find the instructions behind a behavior becomes a useful skill for building and auditing them.
Limits to Respect
SEAT© is for understanding a structure, not for copying content. You reuse a mechanism (outline, tone, argument structure) to produce an original document; you do not reproduce texts, images, or protected elements belonging to others. The method also keeps the user in the role of an editor: they understand how the result is constructed instead of just consuming it.
Choosing the Right Method
| Your Situation | Method | What It Does |
|---|---|---|
| You know exactly what you want | SCOPE© | Frames the request in five axes |
| Your need is still unclear | PROBE© | The AI probes, refines the need, checks the objective, briefs for nuances, then executes |
| The task is too complex for a single request | STEP© | Breaks it down into blocks, tasks, steps, processes |
| You have a model result to reproduce | SEAT© | Finds the prompt that produced it |
| You need to verify an AI's output | SCORE© | Has another AI rate and critique the result |