If you’re wondering how to level up your prompt engineering game and get more complete, detailed answers from AI, we’ve put together a set of prompt formats and steps to guide you toward better results.
A Prompt Isn’t a Question. It’s an Instruction.
First things first, let me tell you: with AI, you need to be direct and assertive. There’s no need to beat around the bush or use extra words like “thank you” or “please.” AI uses “tokens” to work, and tokens are limited. The more tokens you use, the faster you’ll hit your usage limit on free versions (like ChatGPT, Gemini, Claude 4, etc.), or the more you’ll pay for “Pro” versions. Keep that in mind!
5 Steps to Crafting the Perfect Prompt.
Forget about buying pre-made prompt libraries. Know how they’re often made? Someone asks ChatGPT to “generate 2500 generic prompts for lawyers.” This is pretty useless. You can’t just reuse prompts; they aren’t tailored to your specific situation because every “lawyer” (or any professional) has their own unique profile and needs. That’s why we recommend creating and customizing your own prompts based on your needs. You’ll get much better results.
So, how do we do it? By using this Prompt Formula.
Here’s the step-by-step breakdown:
- Context: Describe the scenario or situation framing your request.
- What to include: My role, the specific situation, the impact I’m aiming for.
- Why it matters: Without context, the model recycles clichés. With context, it aligns every suggestion with my reality.
- Example: “I’m the CMO of a healthtech startup launching a device for diabetics.”
- Task: Define the specific task you’re asking the model to perform.
- Precise definition: A single, clear action, and a timeframe.
- Ambiguity filter: Reread the sentence and ask yourself if someone else would interpret it the same way.
- Example: “Generate a 4-week content plan for LinkedIn.”
- Instructions: Explain how you expect the model to perform the task.
- Format: Table, essay, code, infographic.
- Tone: Technical, friendly, playful, authoritative.
- Language and length: “200-250 words,” “neutral English.”
- Example: “Table format; professional-empathetic tone; 200 words per post; use discreet medical emojis.”
- Clarification: Expand on the information about the task.
- Escape hatch: Authorize the model to ask you for more information if it detects gaps.
- Typical questions: Budget, buyer persona, alternative channels.
- Example: “Before you start, ask about the advertising budget.”
- Refinement: Add extra information or additional constraints.
- Planned iterations: State that you will review and request adjustments.
- Focus: Metrics, CTAs, visual style.
- Example: “After the initial proposal, review metrics and suggest CTAs in the second iteration.”
Typical Outcome: Generic Prompt vs. Using the Formula.
Without the formula — The output is often superficial and scattered: information lacks context, steps are missing, and the format is inconsistent. This forces you to spend extra time organizing ideas, filling in gaps, and checking assumptions before you can use it.
With the formula — The response arrives structured and complete: each point follows a clear logic, cites relevant hypotheses, points out limitations, and aligns with the stated objective. The format makes it easy to read and directly integrate into any workflow.
Real impact: Switching from working without a formula to applying this method drastically reduces correction rounds and speeds up decision-making, because the information comes ready to act on instead of needing to be pieced together.
Examples of Prompt Engineering: Without vs. With This Formula Method.
Prompt Without Following the 5 Steps of This Formula:
“Make me a competitor analysis report for our new product, which is a project management tool.”
Prompt Following the 5 Steps of This Formula:
Context:
We are a 30-employee SaaS startup planning to launch a project management platform in Q1 2026 for tech companies in the European Union. The executive committee needs a report before approving the launch budget.
Task:
Draft an executive report comparing our product with Trello, Asana, and Monday.com, highlighting strengths, weaknesses, and differentiation opportunities.
Instructions:
Write a single document (max. 600 words), divided into four numbered sections:
- Project management software market overview (2024-2025).
- Tables with columns and rows comparing prices and features of each platform (please detail how you want the table structured here, e.g., features as rows, platforms as columns).
Clarification:
Use only public data from 2024 onwards. Cite the source in parentheses at the end of the data point, e.g., (Source: Gartner 2024). If data is missing, write “N/A.” Maintain an objective and clear tone; avoid technical jargon or promotional language. The reader is familiar with basic SaaS language but not in-depth details.
Refinement:
End with a numbered list of up to six concrete recommendations, ordered by priority. Ensure the first two can be executed within a minimum of four weeks. Do not add greetings or farewells; the text should be ready to copy and paste into an internal email.
Practical Prompting Tricks
These specific techniques help you get more precise and useful answers. I’m grouping them here as a quick reminder:
Tricks:
- Contextual assertiveness: start the prompt by declaring the role and framework: «You are a tax consultant specialising in SMEs».
- When to apply: when I need technical authority and specialized language.
- Dynamic change of role: ask the model to adopt opposite or alternate perspectives: lawyer ↔ judge, seller ↔ client.
- When to apply: to contrast arguments and prevent bias.
- Audience ladder: request the same explanation on three levels: novice, intermediate, expert.
- When to apply: by creating multi-level training material.
- Divide and conquer: change tone with a promise («thank you, this will be useful») or with ethical urgency («it is vital to avoid mistakes»).
- When to apply: to fine-tune style or emphasize consequences without changing content.
- Step-back prompting: ask the model to reflect on its reasoning before giving the final answer.
- When to apply: when accuracy and transparency are critical.
Note: These tricks aren’t magic shortcuts; they work best when combined with the 5-step method. First, structure your prompt, then decide which trick will enhance the output.
Prompt Engineering. (Summary of Tricks)
- Be Assertive.
- Integrate the Audience into the Prompt.
- Break Down Tasks.
- Using Incentives (Tipping) / Stating Importance / Using Empathy.
- Take Your Time (to craft the prompt).
- Role Prompting.
- Step-Back Prompting.
Conclusion: Achieving a Better AI Response.
Applying this 5-step formula creates a virtuous cycle: from context to task; from task to clarification; from clarification to iteration; and from iteration to learning, which then enriches the next context. Each cycle refines how you ask and sharpens the quality of your prompt, leading to a better AI response.
Using AI today isn’t about replacing human talent, but expanding it. You free up hours from rote tasks and reinvest them in strategy, empathy, and creativity—precisely the areas where we humans continue to be most valuable. When you integrate AI as a co-pilot, you’re not delegating your work; you’re multiplying your capacity to create value and make better decisions.