The 'All-in-One Prompt' Syndrome
When you ask for a complex document all at once, the AI produces text that collapses in length, coherence, or precision. You lose track of where it drifted, and correction means starting over. STEP© does the opposite: you get the outline, correct it, then produce block by block with a satisfactory level of control at each stage.
The Four Levels
- S, Separate. Split the problem into distinct blocks.
- T, Tasks. Identify the tasks inside each block.
- E, Execution. Detail the sequence and stages for each task.
- P, Process. Chain it all into a clear process.
The STEP© Sample Prompt
I have a complex task: [describe the project] Use the STEP© method. Perform a Separation into several Tasks, detail the Execution sequence and stages for each task after asking me all the questions about the information you need, then chain everything into a clear Process. Do not write the content now. First, propose the detailed outline. Wait for my validation between each step.
STEP©: proprietary method of Revenu IA, © Mustapha Benkalfate. Reproduction of the sample prompt is authorized with source attribution.
For Which Deliverables?
- A report or a study of several dozen pages.
- A keynote or a training session, with sequences, exercises, and materials.
- A proposal in response to a call for tenders.
- The design of an agent: sources, instructions, safety rules, tests.
Why Keep the Breakdown for Yourself
Owning the breakdown means retaining the mental model of the project. If the AI breaks down and produces everything, you become what Mustapha Benkalfate calls a 'fragile expert': unable to explain how the parts fit together, and therefore unable to correct the whole. STEP© protects you from this liability: you understand the document because you validated its structure.
STEP© and Agents: The Human as Orchestrator
During a session with an executive committee in 2026, Mustapha Benkalfate presented STEP© as a skill that gains value with AI agents. An agent launched with vague instructions can run for a long time, consume many tokens, and produce little value. With usage-based billing replacing unlimited subscriptions, a poorly broken-down instruction costs money. The human who knows how to break down a task remains the orchestrator.
Choosing the Right Method
| Your situation | Method | What it does |
|---|---|---|
| You know exactly what you want | SCOPE© | Frames the request along 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, execution stages, and a clear process |
| 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 |