Transcription
Most people get shallow, generic answers from AI because they ask it to help without telling it how to think. That's a skill 99% of people are missing. In this video, I'll share a formula that can take your prompting skills from beginner to advanced. I call it CRAFT. It might look like a prompt formula, but it's actually a thinking framework. Once it trains your brain to think systematically while unlocking AI's full reasoning power, by the end of this video, you'll walk away with three things: better AI responses right away, the technical understanding that actually matters, and most importantly, a thinking framework that makes you better at problem-solving, not just prompting. Because in an AI-driven world, that's the real competitive advantage.
You've probably heard technical terms like token probability, context windows, fine-tuning, and so on. They're useful knowledge, sure, but they won't immediately transform your prompting. If you want to go from beginner to advanced quickly, there's one technical concept that matters the most: pattern recognition. This is how AI fundamentally works with your input. Get this right, and the rest becomes so much easier.
Think of AI's training data as a massive library filled with millions of tagged and categorized conversations. When you make a request, the AI searches for similar patterns it has seen before. Here's the key: a generic prompt like "help with a presentation" only activates surface-level patterns, like pulling a plain manila folder off the shelf. But a specific prompt like "help me explain budget delays to stressed executives" activates multiple, more precise patterns that combine to create something tailored. It's like mixing paint colors. The more specific colors you choose, the richer and more useful your result becomes.
So, how do you consistently give AI the right clues to find the best patterns? That's exactly what CRAFT helps you do. CRAFT stands for Context, Role, Action, Format, and Thinking Mode. Each element taps into AI's pattern recognition in a different way. You mix and match them depending on what you need. Some elements, like Context and Role, are useful in most situations. Others, like Action, Format, and Thinking Mode, are more specialized. You pull them in when you need sharper outcomes, a specific structure, or a more strategic approach. Let's break them down one by one.
Starting with Context and Role. Context is the foundation. It's about painting the picture, giving it the background it needs to understand what you're really dealing with. That's how you define specific patterns instead of generic ones. You already do this without thinking. Anytime you explain a task to a coworker or a classmate, you start by describing a situation. Same thing with AI. That's your starting point.
Here are some key context categories that tend to work well: the situation, the people, the limits, and the past. Notice what's happening here: you're not just dumping information randomly. You are systematically mapping the landscape of your problem. The situation gives AI the "what," the people give it the "who," the constraints give it the limitations, and the background gives it the "why it matters." Now, this is exactly how strategic consultants tackle complex problems. They map the terrain before they design a strategy. You don't need all four every time. Even one or two will make your prompt much stronger.
Let me give you an example. Instead of saying "help me plan a party," add details about the situation, the people, the limits, and the past. Now AI knows it's not just any party; it's your actual challenge with real constraints, and the output changes completely. AI is now matching against patterns like "kids birthday party," "budget constraints," "indoor space," not just a generic "party planning" bucket.
Now let's move on to Role. This is how you tap into AI's expertise library. You're defining both the kind of expert you need and who they should be communicating with. That way, AI pulls from the right professional communication patterns. Instead of falling back on generic responses, you wouldn't ask your accountant for relationship advice or your therapist for tax strategy. You pick the right role depending on the problem. Same idea with AI.
There are four role categories that work quite well: expertise or skills, domain or field, audience, and approach or style. Role has a dual nature that most people miss. You're not just telling AI what expert to be; you're also defining the audience. Here's how it works: expertise sets the knowledge base AI pulls from. Domain gives it the industry context. Audience shapes the communication style. Approach defines the relationship dynamic. When you add these details, AI activates the right conversation patterns instead of giving you one-size-fits-all answers.
Here's how Role transforms a prompt. Instead of asking something broad like "help me improve my side hustle idea," specify the expertise, the experience level, and the way you'd like it to communicate. Now, AI pulls from entrepreneurial patterns, not just generic business advice. The result is practical, experience-based guidance that's tailored to a beginner's mindset, instead of abstract theories or frameworks.
Now, we move into the advanced elements, starting with Action. This is where you show problem-solving clarity. It helps you to make wishes like "give me some ideas" into specific, measurable outcomes, and it helps AI focus on actionable solutions instead of wandering into generic advice. Think about it like hiring someone to clean your house. If you just say "clean the kitchen," they might wipe the counters but ignore the oven or drawers—the spots that actually matter to you. The clearer you are, the better the result. Same with AI.
Action breaks into four clear types that eliminate confusion: the output, the impact, the choice, and the fix. You either need a specific output or deliverable created. You want to achieve a particular impact or outcome. You need help making a decision or choice, or you have a problem that needs a solution or fix. Here's a big mistake people make: they assume "help" is an action. It isn't. It's just a wish. Real actions use specific verbs with clear boundaries: create, evaluate, identify, compare. Notice how each type requires a different kind of thinking from AI. Deliverables call for creation skills. Outcomes require strategy. Decisions need analysis, and solutions demand problem-solving.
Let me give you an example. Instead of "help me get fit," try building a prompt around these categories: what to produce, what decision to make, what problem to solve, and what capability to build. Now AI can deliver specific workout plans, problem anticipation, and progress tracking. You're guiding it toward exactly what you need, not just what sounds vaguely helpful.
Format is all about making your output useful. You're structuring AI's responses so it matches exactly how you will use the information. This helps AI pick the right template from its pattern library instead of guessing or providing random structure. Think about how you do this naturally. When you ask someone to write something for you, you usually specify the format. A shopping list looks different from a presentation outline, which looks different from step-by-step instructions.
To guide AI effectively, focus on four format categories: structure, tool, reference, and vision. Format isn't about being fancy; it's about being functional. Structure determines how information flows. Media matches your platform needs. Reference gives a proven template to follow, and visual considerations ensure the result is easy to use. Form follows function. The format should always serve your purpose, not just look good. Every time you practice this, you're learning to match structure to intent, a skill you can transfer to any kind of communication.
Here's an example. Instead of a vague request, give AI a prompt with formatting details. Ask for a step-by-step checklist with prep times, a printable shopping list organized by store sections, quick reference cards, and timing critical steps highlighted. Now, AI follows specific document patterns instead of choosing randomly. The result is output you can immediately use: clear, organized, and actionable.
Here's where you move into the advanced level: choosing cognitive approaches. Thinking Mode is your strategic move. You're telling AI not just what to do but how to approach the problem mentally. It helps AI switch between different reasoning patterns, like calling a different expert consultant from its training data. This is what separates beginners from experts. Beginners just ask AI to help, while experts direct AI's thinking process.
There are seven core thinking modes: storytelling, diagnostic, systematic, creative, persuasive, reasoning, and empathetic. Each mode activates a different reasoning pattern in AI: storytelling for engagement, diagnostic for problems, systematic for complexity, creative for innovation, persuasive for influence, reasoning for logic, and empathetic for relationships. For simple tasks, AI often picks up the right mode automatically. For example, if you ask for an apology email, it naturally leans empathetic. But for complex challenges or strategic decisions, being intentional about your mental approach makes all the difference.
For example, say your task is "help me communicate with a difficult colleague." With diagnostic thinking, you focus on identifying the root cause of the conflict. With persuasive thinking, you focus on motivating better collaboration. With empathetic thinking, you aim to understand their perspective and find common ground. Same challenge, different thinking modes, and each calls for a distinct strategy. This is cognitive flexibility in action.
Thinking Modes are also a crucial part of context engineering, the practice of creating the best information environment for AI to operate in. There's a lot more to explore here, but I'll save that for another video. For now, just being aware of these modes and intentionally selecting them when tackling complex tasks gives you a huge advantage. You're not just getting AI responses; you're directing AI how to think.
Now, let's see how this works in practice. Everything I've explained might look like a prompt formula, but it's really a thinking framework. It helps you ask the right questions up front, keeping you in control of the AI conversation. Here's a real example: preparing a competitive analysis deck. I'll walk you through three versions of the same request to show how thinking and results evolve.
Prompt one just states the task directly to ChatGPT. Prompt two adds some context. Most people operate at this Prompt Two level. You know, specificity helps, but you are not approaching it systematically. Notice the difference in AI's responses. With Prompt One, AI asks standard clarifying questions that aren't tailored to a situation. Prompt Two, the questions become more specific and strategic. The follow-up questions in Prompt Two are better.
But here's the problem: most people miss when your initial prompt is incomplete. AI ends up leading the conversation for complex challenges. This can create mental frustration because you are spending cognitive energy answering AI's agenda instead of moving toward your specific needs.
Let's apply the CRAFT framework and see how it systematically helps us gather the right information before you even touch AI. CRAFT can guide you through these questions: What's my situation? You realize you need to clarify the timeline with your manager and understand why leadership is concerned. Who's involved? You identify your audience preferences and decision-making style. What do I need? You define specific deliverables and success metrics. How should it be delivered? You determine the format that works best for your executives. This preparation process ensures you collect complete information and create prompts that work the first time.
Watch what happens. ChatGPT begins by analyzing your request with pattern recognition and diagnostic reasoning, identifying the core problem and context. From there, it structures insights into a tailored response, organized, actionable, and aligned to your goals, while leaving room for you to guide or adjust the direction. This creates a strong foundation and ensures you maintain control over what matters the most.
This example shows two fundamentally different interaction styles with AI: reactive versus strategic. Whatever challenge you're tackling, staying proactive and in control always produces better results than letting AI lead the conversation. I've also created a CRAFT cheat sheet you can use to quickly identify which elements might be missing in your prompts. You'll find the link in the description below. If this helps you see prompting differently, please subscribe and like the video. It encourages me to create more content like this.
Build cognitive assets, not prompt collections. In this video, you weren't just learning how to prompt better; you are learning how to think better. And every time you prompt, you're not just getting better AI responses; you are developing systematic thinking. This isn't prompt engineering; it's cognitive engineering. That's how you get ahead of 99% of people who only focus on prompting techniques. I also have another video sharing one simple practice that will develop your intuition for solving problems with AI.