My AI Prompt Blueprint - Steal it And Get Great Results Every Time

The Concept: Crafting a Prompt

It’s 4pm on Friday.

The deadline is in an hour.

You race to your computer, open up ChatGPT, tell it what you want, and press enter.

(waiting)

ChatGPT starts writing your content and… it’s mediocre. It’s missing some things, and some of the data is wrong.

So you tell ChatGPT that it’s an idiot (just joking; never say this because when AI takes over, it’ll remember how you treated it) and it needs to try again. You type more directions into the chat box, press enter, and it’s still not right.

(50 minutes later)

After more back and forth with ChatGPT, you finally get the output you want. It looks good, everything seems to be in place, and the numbers appear correct.

(fast forward a week)

It’s 4pm on Friday… again.

The deadline is in an hour… again.

You race to your computer, open up ChatGPT, and spend 50 minutes in back and forth with it… again.

Sounds like Bill Murray in Groundhog Day, right?

Let me tell you, there’s an easier way.

The easier way is to build a prompt that you can reuse again and again.

I do it all the time for my businesses, and other expert prompt engineers do it too. It’s how you save time and get incredible results from the AI.

Let me share with you the seven components that make up a reusable prompt.

The 7 Components of a Prompt

I focus on seven key components when writing a prompt that I want to save and reuse at a later point in time. You don’t need every component all the time, but each one has its own job and helps solve a specific problem.

Once you get comfortable with these seven parts, you’ll have everything you need to create prompts you can reuse in the future. Best of all, they’ll give you professional results every time.

1. Instructions (The Mandatory Core)

This is what every prompt needs most: a clear action verb and task that tells the AI exactly what to do. You’ve got to spell out what you want, crystal clear, every single time you write a prompt.

What is it: The primary command or question that usually consists of a clear action verb and a task that tells the AI what to do.

When to use it: Always.

Example #1:

AI Prompt

Write a Facebook post about our Grill Knight brush which is a chainmail brush for scrubbing grill grates.

The action verb is “write” and the task is writing a “Facebook post about our Grill Knight brush…”.

Example #2:

AI Prompt

Write an email to a customer to leave a review for our Grill Knight brush which is a chainmail brush for scrubbing grill grates.

The action verb is “write” and the task is writing an “email to a customer to leave a review…”

2. Role (The Expert Hat)

The next part of my prompting blueprint is the role, and it’s optional just like the other components below. You don’t need to use it unless it helps your prompt.

A role lets you tell the AI who it should pretend to be. When you give the AI a role, it’s like instantly hiring a top expert in that field. The AI will start using specific knowledge, vocabulary, and thinking patterns that match that profession.

You might want the AI to act like a famous person (e.g., Act like Albert Einstein), but that’s not what the role section is for. I’ll explain more in the “style” section below.

What is it: An instruction that assigns a persona or job title to the AI (e.g., “Act as a…”).

When to use it: When you need specialized knowledge or a specific point of view.

Example #1:

AI Prompt

Act as an expert social media manager and write a Facebook post about our Grill Knight brush which is a chainmail brush for scrubbing grill grates.

Telling AI to act as a social media manager might unlock specific knowledge that’ll help it write better Facebook posts.

Example #2:

AI Prompt

Act as a customer success manager, and write an email to a customer to leave a review for our Grill Knight brush which is a chainmail brush for scrubbing grill grates.

In this example, telling the AI to act as a customer success manager does way more than just saying “social media manager” like in the previous example.

Why?

Because “customer success manager” gives you a specific viewpoint. Think about it: if I asked the AI to write that email as the “founder of Grill Knight,” I’d get a totally different angle.

But with “social media manager,” it’s pretty generic. It’s almost too broad and doesn’t really add much.

So the takeaway is to test how the AI writes with the role versus without it. Use whatever looks better in your opinion.

3. Background (The Secret Knowledge)

The background component is something I use a lot when I prompt. It’s for giving crucial info that the AI usually doesn’t know about. This includes specific context about a person, company, product, or other thing that AI doesn’t have knowledge about.

What is it: Information, not instructions. It’s the context the AI needs to do its job properly. Usually it’s information the AI doesn’t know.

When to use it: When the AI needs specific information to complete the task accurately that it doesn’t know about.

Example #1:

AI Prompt

Act as an expert social media manager and write a Facebook post about our Grill Knight brush which is a chainmail brush for scrubbing grill grates. It is made from 316-grade marine stainless steel with 3,800 interlocking rings in a bristle-free chainmail design, this 7″ x 5″ scrubber safely cleans all grill types without scratching while featuring a heat-resistant silicone handle. Dishwasher safe and backed by a lifetime warranty, it weighs just 14 ounces for easy handling and will never shed dangerous wires into your food.

In this example, we’ve added some specific details about our product, the Grill Knight brush, that the AI would not know about.

Example #2:

AI Prompt

Act as a customer success manager, and write an email to a customer to leave a review for our Grill Knight brush which is a chainmail brush for scrubbing grill grates. It is made from 316-grade marine stainless steel with 3,800 interlocking rings in a bristle-free chainmail design, this 7″ x 5″ scrubber safely cleans all grill types without scratching while featuring a heat-resistant silicone handle. Dishwasher safe and backed by a lifetime warranty, it weighs just 14 ounces for easy handling and will never shed dangerous wires into your food.

Notice how our background information does not include directions. It’s just information that the AI needs to know to do its job when writing about the subject.

4. Section Variables

I use section variables in my prompts almost as much as the background component we just discussed.

You might hear others call these delimiters, but I think the phrase “section variables” captures their usefulness perfectly.

I like to use them to section off different pieces of information in my prompt so the AI doesn’t get confused. I can also refer back to these sections in my directions or instructions by using the variable name, just like someone would do when programming.

What is it: A user-defined label for a specific chunk of information within a prompt.

When to use it: When your prompt requires multiple pieces of context, and you need to tell the AI how to use each specific piece.

Example #1:

AI Prompt

[PRODUCT DETAILS]:
The Grill Knight is a chainmail grill brush for scrubbing grill grates. It is made from 316-grade marine stainless steel with 3,800 interlocking rings in a bristle-free chainmail design, this 7″ x 5″ scrubber safely cleans all grill types without scratching while featuring a heat-resistant silicone handle. Dishwasher safe and backed by a lifetime warranty, it weighs just 14 ounces for easy handling and will never shed dangerous wires into your food.

[INSTRUCTIONS]:
Act as an expert social media manager and write a Facebook post about our product using the information in the [PRODUCT DETAILS].

In this example, I haven’t added new directions or background. I’ve just moved around some of the information using section variables, which makes things clearer for both me and the AI.

When you create separate sections for some of your data, the AI won’t get confused as easily. Plus, it makes swapping out data in your prompt super easy and efficient.

Notice that the variables I used make contextual sense from a naming perspective and I’ve also encased them in square brackets. Using square brackets to encase section variables along with uppercasing the variable name itself is common practice.

But regardless of how you format your variables, the key is consistency. You don’t want to use one style for one [VARIABLE] and another style for another {variable}. This can confuse the AI, causing it to spend unnecessary mental effort that could’ve been directed at following your directions and crafting a higher quality result.

Example #2:

AI Prompt

[PRODUCT DETAILS]:
The Grill Knight is a chainmail grill brush for scrubbing grill grates. It is made from 316-grade marine stainless steel with 3,800 interlocking rings in a bristle-free chainmail design, this 7″ x 5″ scrubber safely cleans all grill types without scratching while featuring a heat-resistant silicone handle. Dishwasher safe and backed by a lifetime warranty, it weighs just 14 ounces for easy handling and will never shed dangerous wires into your food.

[INSTRUCTIONS]:
Act as a customer success manager, and write an email to a customer to leave a review for our product using the information in the [PRODUCT DETAILS].

See how easy it is to reuse this prompt with different product details or instructions? Just swap out what you need and you’re good to go!

5. Style (The Voice and Vibe)

A style component in your prompt lets you control the personality of your output, and it has three parts: tone, audience, and persona.

I don’t have a specific format or standard way to add style, but I’ll usually start with “write in a style that is…” (for tone), and I might add “for an audience of…” (for audience).

Sometimes I’ll also include “while writing, write like…” (for persona).

Most often though, I just use tone; for example, I’ll say “write in a style that’s friendly, encouraging, and appreciative.”

The real secret is to test this a lot and see how it changes your output. If you want to use a style component in your prompts, try writing a few versions without it first, then try some with it. You’ll quickly see the difference it makes.

What is it: Instructions that define the emotional feel, target reader, and writing style.

When to use it: When the feeling and reception of the message are as important as the content itself.

Example #1:

AI Prompt

You can go as simple or detailed as you want with the style. But no matter how much detail you use, make sure you test the output first. The style you add will have the biggest impact on how readable the AI’s output is overall for your audience.

Example #2:

AI Prompt

[PRODUCT DETAILS]:
The Grill Knight is a chainmail grill brush for scrubbing grill grates. It is made from 316-grade marine stainless steel with 3,800 interlocking rings in a bristle-free chainmail design, this 7″ x 5″ scrubber safely cleans all grill types without scratching while featuring a heat-resistant silicone handle. Dishwasher safe and backed by a lifetime warranty, it weighs just 14 ounces for easy handling and will never shed dangerous wires into your food.

[INSTRUCTIONS]:
Act as a customer success manager, and write an email to a customer to leave a review for our product using the information in the [PRODUCT DETAILS]. Write in a style that is playful, sizzling, and grate-ful.

Another thing you might’ve noticed is that I added the style details into the [INSTRUCTIONS] variable. This was my personal preference, but there’s no need to create separate variables for every last rule or direction you give the AI as long as everything is clear.

6. Limits (The Rules and Guardrails)

The limits component is something I use a lot when I prompt. It’s where you set the boundaries for the AI’s response. This covers the output format, length constraints, and any specific “dos and don’ts.”

What is it: A set of rules that constrain the output.

When to use it: When the final output needs to fit a specific size, structure, or include/exclude specific elements.

Example #1:

AI Prompt

If you’ve been paying attention, then you could technically argue that “Facebook post” is also a limit since it tells you what format to use for your output (e.g., whatever the AI thinks a Facebook post should look like).

However, something like this is usually part of the original prompt in the instructions component, the first one I talked about. So whether we consider it part of the limits component is merely semantics.

Example #2:

AI Prompt

[PRODUCT DETAILS]:
The Grill Knight is a chainmail grill brush for scrubbing grill grates. It is made from 316-grade marine stainless steel with 3,800 interlocking rings in a bristle-free chainmail design, this 7″ x 5″ scrubber safely cleans all grill types without scratching while featuring a heat-resistant silicone handle. Dishwasher safe and backed by a lifetime warranty, it weighs just 14 ounces for easy handling and will never shed dangerous wires into your food.

[INSTRUCTIONS]:
Act as a customer success manager, and write an email to a customer to leave a review for our product using the information in the [PRODUCT DETAILS]. Write in a style that is playful, sizzling, and grate-ful. Write under 100 words.

When you’re setting an output limit on your content, “words” or “paragraphs” work best. I’ve noticed the AI really struggles when you try to limit output to a certain number of characters (e.g., limit the output to 63 characters).

Here’s another thing I’ve found: when you limit the output to a specific length, the quality can drop dramatically. So I’d recommend doing some tests both with and without a limit to make sure there isn’t a huge drop in quality.

7. Examples (The Show Don’t Tell Trick)

The examples component of my prompt blueprint is something I used to use all the time. I don’t use it as much anymore, but it’s still very important in certain use cases.

Why is it so powerful?

Because it’s one of the fastest ways to teach the AI your style or format. When you give a clear example of what you want, you’re giving the AI a perfect model to copy. Another name for this is few-shot prompting.

What is it: A sample of the desired output you want the AI to generate.

When to use it: When you need a very specific, nuanced style or structure that is hard to describe with words alone. You should also use it when the AI isn’t following your output formatting directions.

Example #1:

AI Prompt

I didn’t type out all the examples to save space, but you get the idea.

When you’re giving examples, the most important thing is focusing on what actually matters. AI learns through patterns, so if your examples don’t have a clear pattern, you can’t expect the AI to give you what you want.

What makes a pattern? Well, it depends on what you’re asking for as output. In this case, we’re asking for a Facebook post. Maybe you picture a Facebook post as having three paragraphs, then a call to action, then three hashtags, and no emojis. If you give two or three examples of this, the AI can follow along.

Now, when making a Facebook post, you really don’t need examples. You can simply be clearer with your instructions. But you might very well run into situations where you’re creating some kind of output in a specific format, or with specific details, which are hard to describe in the instructions and you’re not getting consistent output from the AI. In that case, using examples can make a huge difference.

Example #2:

AI Prompt

[PRODUCT DETAILS]:
The Grill Knight is a chainmail grill brush for scrubbing grill grates. It is made from 316-grade marine stainless steel with 3,800 interlocking rings in a bristle-free chainmail design, this 7″ x 5″ scrubber safely cleans all grill types without scratching while featuring a heat-resistant silicone handle. Dishwasher safe and backed by a lifetime warranty, it weighs just 14 ounces for easy handling and will never shed dangerous wires into your food.

[INSTRUCTIONS]:
Act as a customer success manager, and write an email to a customer to leave a review for our product using the information in the [PRODUCT DETAILS]. Write in a style that is playful, sizzling, and grate-ful. Write under 100 words.

[EXAMPLE 1]:
(an ideal output goes here)

[EXAMPLE 2]:
(an ideal output goes here)

[OUTPUT]:

You don’t have to use section variables to separate your examples. But whatever you do, just make sure the AI can understand where each example starts and stops.

Wrap Up

The seven components above in my prompting blueprint can make you a world class prompter almost overnight.

But here’s the thing: the secret to high-performance prompting isn’t cramming all seven of these components into every request. It’s about looking at your task and asking, “Which components will have the biggest impact here?”

For a quick brainstorm, instructions might be enough.
For a critical piece of ad copy, you’ll likely need instructions, role, style, and limits.
For analyzing a meeting transcript, you’ll need instructions and background.

Mastering AI is like any other craft: it’s about learning your tools and developing the wisdom to know when and when not to use them.