Once you understand the basics of prompt engineering, the next step is building a toolkit of proven prompt formulas you can reuse. Instead of starting from scratch each time, you can adapt these templates for your specific needs. Each formula below is battle-tested and works across ChatGPT, Claude, Gemini, and most other text-based AI tools. Let’s dive in.
Formula 1: The Role-Context-Task-Format Framework
This is the most versatile prompt structure and the one you’ll use most often. It works for almost any task:
Template: “Act as [role]. I’m [context about you/your situation]. I need you to [specific task]. Format your response as [desired format].”
Example: “Act as an experienced career coach. I’m a marketing manager with 5 years of experience looking to transition into product management. I need you to help me identify transferable skills and gaps in my resume. Format your response as a two-column table: one column for skills I have, one for skills I need to develop.”
Why this works: it gives the AI a clear perspective to adopt, provides context about your situation, defines the task precisely, and specifies the output format. The AI has everything it needs to give you a useful, structured response.
Formula 2: The Question-First Approach
Sometimes you don’t know what information the AI needs to help you. Instead of guessing, ask the AI to interview you first:
Template: “I want help with [general goal]. Before responding, ask me 5-7 questions that will help you understand my situation better. Wait for my answers, then provide your response.”
Example: “I want help creating a personal budget. Before responding, ask me 5-7 questions that will help you understand my financial situation and goals. Wait for my answers, then create a detailed monthly budget for me.”
Why this works: it surfaces information you might not have thought to provide, leading to much more tailored and useful responses. The AI essentially does the prompt engineering for you.
Formula 3: The Multiple Options Request
When you’re not sure exactly what you want, asking for multiple variations helps you explore possibilities:
Template: “Create [number] different [type of content] for [purpose]. Each option should [varying criteria]. Label each option clearly.”
Example: “Create 5 different email subject lines for a newsletter about productivity tips. Each should have a different tone: 1) direct and benefit-focused, 2) curiosity-driven, 3) question-based, 4) story-based, 5) bold and provocative. Label each option clearly.”
Why this works: it forces the AI to explore different angles rather than settling on its first idea. You get variety and can mix and match elements from different options.
Formula 4: The Step-by-Step Decomposition
For complex tasks, asking the AI to break things down step by step improves quality:
Template: “I need to [complex task]. Break this down into clear steps, and for each step, explain what to do and why it matters. Think through this step by step before providing your final answer.”
Example: “I need to plan a 3-day conference for 200 attendees. Break this down into clear steps covering everything I need to do from 6 months out to the day of the event. For each step, explain what to do and why it matters. Think through this step by step before providing your final answer.”
Why this works: “thinking step by step” is a proven technique that helps AI (and humans) tackle complex problems more effectively. You get more thorough, well-organized responses.
Formula 5: The Critique and Improve Loop
Instead of accepting the AI’s first response, ask it to critique and improve its own work:
Template: First prompt: “[Your request].” Second prompt: “Critique the response you just gave. Identify three weaknesses or areas for improvement, then provide an improved version.”
Example: First prompt: “Write a 200-word introduction for a blog post about why beginners should learn to use AI tools.” Second prompt: “Critique the introduction you just wrote. Identify three weaknesses (such as vague claims, weak hook, or unclear value proposition), then provide an improved version.”
Why this works: AI can often identify weaknesses in its own work when asked to critique it. The improved version is typically stronger than the original.
Formula 6: The Persona Interview
This is perfect for brainstorming, getting diverse perspectives, or role-playing scenarios:
Template: “Adopt the persona of [specific type of person]. I’m going to [describe the scenario]. Respond as this persona would, including their perspective, concerns, and communication style.”
Example: “Adopt the persona of a skeptical CFO at a mid-sized manufacturing company. I’m going to pitch you on investing in AI training for our customer service team. Respond as this persona would, including their concerns about cost, ROI, and implementation challenges. Ask me tough questions one at a time.”
Why this works: it forces the AI to take a specific viewpoint, which makes its responses more concrete and useful. Great for rehearsing difficult conversations or stress-testing ideas.
Formula 7: The Reverse Engineer
If you have an example of what you want but can’t articulate why it works, ask AI to reverse-engineer it:
Template: “Here’s an example of [type of content]: [paste example]. Analyze what makes this effective, then create a new piece following the same principles for [your specific topic].”
Example: “Here’s an example of a great product description: [paste description]. Analyze what makes this effective (tone, structure, specific techniques), then create a new product description following the same principles for my product: a wireless charging pad made from sustainable bamboo.”
Why this works: examples communicate more clearly than abstract descriptions. By having AI analyze what works, you get a tailored response that captures the qualities you admire.
Combining Formulas for Advanced Use
Once you’re comfortable with these formulas, you can combine them. For example, you might use the Question-First approach to gather information, then the Role-Context-Task-Format framework to specify your actual request, then the Critique and Improve loop to refine the response. This kind of multi-step prompting is how power users get exceptional results from AI.
Don’t feel you need to use complex prompts for everything. Simple, clear prompts are often enough for simple tasks. Save the more elaborate formulas for tasks where the quality of the output really matters.
Building Your Prompt Library
As you use these formulas, you’ll find yourself adapting them to your specific needs. Keep a document or note where you save prompts that worked well. Over time, you’ll build a personal library of prompts you can quickly adapt for new situations. This library becomes one of your most valuable productivity assets, something you can carry with you across jobs and industries.
The goal isn’t to memorize formulas; it’s to internalize the principles behind them. Once you understand why these patterns work, you’ll naturally write better prompts even when you’re not following a specific template. Practice with one or two of these formulas this week, and you’ll be amazed at how quickly your AI results improve.