The Art of Asking

The Art of Asking: Why Prompt Engineering Matters More Than You Think

Vladimir Lenin once said, "There are decades when nothing happens, and there are weeks when decades happen." I believe that we are living in one of those weeks right now, when decades are happening.

If you haven't experimented much with AI tools yet, it is genuinely hard to overstate how capable Artificial Intelligence has become at performing tasks that typically require human intelligence. But the difference between someone who gets mediocre results from AI and someone who gets genuinely useful results often comes down to a single skill: prompt engineering.

What Is Prompt Engineering?

The key to getting good results from AI is learning how to ask good questions. In AI terminology, the questions or instructions you give are called "prompts," and learning to write effective prompts is called "prompt engineering."

Here's the fundamental principle: The more specific and detailed your prompt, the better the result.

Let's say you want help planning a vegetable garden. A basic prompt might be: "What should I plant in my garden?"

That's not a bad start, but the AI doesn't know much about what you want. It doesn't know where you live, what time of year it is, how much space you have, or what your goals are.

The AI has to guess what you really want to know. And AI, like your Uncle Bob at Thanksgiving, will happily talk for twenty minutes about something you weren't actually asking about.

A much better prompt would be: "I live in Northern California (Zone 9) and want to start a small vegetable garden in March. I have a 10x10 foot raised bed with full sun. I'm a complete beginner and want to grow vegetables my family will actually eat. What are the easiest vegetables to start with for this climate and season? Please give me specific planting instructions and a simple maintenance schedule."

Do you see the difference? The second prompt gives the AI your location and climate zone, your timeline, your space constraints, your skill level, your goals, and exactly what kind of information you need back. The more detail you provide, the more useful the result will be.

The Five-Box Framework

Now depending on the complexity of the question you're asking, that determines how much detail you need to include. But if you want the best possible response, especially for more complex questions, here are five things to consider—five boxes to fill in:

Box 1: The Role
Tell the AI what its role is. This is who you want it to be. For example:
"You are an experienced medical professional."
"You are a master gardener with 40 years of experience growing vegetables in Northern California."
The AI will assume that persona. It will draw on everything that someone like that would know and say, "Okay, that's who I am here."

Box 2: The Context
Context is basically everything you can tell it about what's going on. Give it the background, the situation, the relevant details. For example:
"I just got back from the doctor's office, and they told me these are my lab results. There's a possibility that I might need heart surgery. I am a 46-year-old man living in Northern California."
"I want to start a vegetable garden but have never done it before. Last year I killed a cactus, so we're starting from a low baseline here."
This is where you provide the relevant information the AI needs to give you a useful response.

Box 3: The Task
This is what you're actually asking it to do. This is the part that takes a little bit of thought. What do I want to get back from this? What specifically am I asking for?
"What vegetables should I plant in March, and how should I prepare my soil?"
"Can you give me the pros and cons of this treatment versus that treatment?"
Be specific. "Help me with gardening" is less useful than "Tell me which three vegetables are hardest to kill."

Box 4: What You Don't Want
These are the restrictions or the restraints. Tell the AI what you're not looking for. For example:
"No medical jargon or diagnosis."
"Don't ruin the story for me."
"I'm not looking for a 40-page analysis."
This helps the AI avoid going down paths that won't be helpful to you.

Box 5: The Output Format
What do you want it to look like when it's done? Am I looking for a list of bullet points? Am I looking for an essay? What exactly am I looking for?
"I want a fourth-grade level explanation."
"Keep it under 150 words."
"I want 500 words or less."
When you fill in these five boxes, you will get the best possible results. You don't have to do this for everything you ask an AI. If you take a picture of a tomato plant and ask what's wrong with it, you don't need all this detail. But if you need to get the best possible answer for something more complex, this is a good template to work with.

Putting It All Together

Here's an example of how this works:

Basic prompt: "Tell me about Nepal."

If that's your prompt, the AI has to guess what you really want to know. It might tell you about the weather of Nepal. It might tell you about the politics in Nepal. It might tell you about the indigenous people groups in Nepal. It might tell you about Himalayan snow leopards. You're getting something about Nepal, but who knows if it's what you actually needed.

Advanced prompt using the five-box framework:

Role: "Act like an experienced tour guide who's been leading trips through Nepal for the last 30 years."

Context: "You're talking to a person from rural California who's never been out of the state but is considering a trip to Nepal."

Task: "Help this person come up with an itinerary and tell them about the things they would probably enjoy the most."

What you don't want: "No ocean liners to get there—I want to fly."

Output: "Tell me about this in 500 words or less."
When you ask it like that, you're filling in five boxes, one at a time, to get the best possible answer.

Additional Examples

For medical clarity:

• Role: You are a medical professional
• Context: You're speaking to an older audience looking for medical clarity
• Task: Explain the purpose of a cholesterol pill called a statin
• What you don't want: No medical jargon or diagnosis (I don't know what "hyperlipidemia" means and I don't want to)
• Output: I want a fourth-grade level explanation and one question I should ask my doctor about this

For reading guidance:

• Role: You are an expert literary critic who has been studying English literature for 50 years
• Context: I'm reading Dr. Jekyll and Mr. Hyde for the first time
• Task: Tell me what literary elements I should watch for as I read
• What you don't want: Don't ruin the story for me (I know there's something about Jekyll and Hyde, but that's about it)
• Output: Give me 5-7 specific things to watch for, each with a brief explanation

For homeschool planning:

• Role: You are an experienced homeschool teacher who specializes in teaching history to elementary-age children
• Context: I'm teaching American history to my 4th grader who thinks history is boring
• Task: Create a one-week unit study on the American Revolution
• What you don't want: Don't make it too complex—keep activities simple and hands-on. No multi-step craft projects requiring supplies I don't have.
• Output: Include daily activities, discussion questions, and one hands-on project

The Iterative Approach

Here's something else important to understand: don't expect perfection on the first try. Think of working with AI as having a conversation rather than making a single request. The AI won't be offended if you ask it to try again. It doesn't have feelings…yet…probably…we think.

If the first response isn't quite what you want, refine your prompt:

• "That's good, but can you make it more concise?"
• "Can you rewrite that for a younger audience?"
• "That's too formal. Can you make it more conversational?"
• "Can you add a specific example to illustrate this point?"

You can keep refining until you get exactly what you need. This iterative process is actually faster than trying to craft the perfect prompt from the beginning. Start with a solid attempt, then adjust based on what you get back.

Moving Forward

The best advice I can give you about working with AI is to simply try it. Pay attention to how you ask your questions. Use the five-box framework when appropriate:

1. Role - Who should the AI be?
2. Context - What's the background?
3. Task - What specifically do you need?
4. Restrictions - What don't you want?
5. Output - What format do you need?

You might be surprised at how useful it becomes.