Write Better Prompts With ChatGPT

I am so sorry, I was making some last-minute changes to this week’s newsletter and the last edition had a few things I wanted to clarify before sending, I apologize for the double send this week. Please forgive the mistake. It’s mostly the same content, but a little more clarification on the final example, as it was originally going to be a prompt of the week for my Friday send.

Sorry for the spam, Mark 

Everyone’s using AI.

But I notice that many aren’t getting repeatable results.

That’s because most users treat prompts like one-off Google searches—typed once, forgotten the next day.

But good prompts aren’t disposable. They’re assets.

Here’s how to use ChatGPT to not only write better prompts—but to refine, test, and save them into reusable templates for marketing, ops, sales, and more.

Write Better Prompts With ChatGPT

You can use ChatGPT to create prompts for complex tasks and make them reproducible.

Most people write prompts in a rush and get whatever result the model spits back. Then they retype a new version next time—forgetting what worked, what didn’t, and why.

What’s most common is called zero-shot prompting: you ask the model to do something with no examples or context. It gives you a result but with little direction.

Examples:

“Write a cold outreach email for my company.”
“Summarize this article.”
“Write a business plan for my company.”

You might get lucky. But more often, you end up rewriting or asking again.

The key to quality, consistency, and scale is designing prompts intentionally—and using few-shot or many-shot examples to steer the model.

Zero-Shot vs Few-Shot vs Many-Shot: A Quick Primer

More examples = more control. You’re showing the model the rules and the style—so it can generalize better.

Step-by-Step: How to Use ChatGPT to Create Prompts

My prompt engineering practice has evolved significantly over time. Initially, I wrote prompts directly in Notion, using it as both a workspace for iteration and a repository for storage. This approach worked but had limitations in terms of feedback and refinement.

Today, my workflow is more sophisticated. I leverage ChatGPT as a collaborative partner in crafting high-quality prompts before archiving them in Notion for long-term reference and reuse. The technique I've found most valuable is example-driven prompt engineering.

When I identify a task I perform repeatedly, I collect successful examples of both inputs and outputs. I then use these examples to reverse engineer effective prompts by identifying patterns and extracting the critical elements that produce the desired results. These examples also serve as guidance for the AI, providing clear demonstrations of my expectations. This methodology has several advantages:

It grounds the AI in concrete patterns rather than abstract instructions It reduces ambiguity by showing rather than telling It creates consistency across multiple interactions with different AI systems It allows for systematic improvement over time as I collect more examples

Here’s the process for you to follow.

1. Use ChatGPT to Reverse Engineer a Prompt

Provide an example of something you like then ask ChatGPT to create a prompt to duplicate the result.

Initial Input:

Perhaps you want to summarize our new feature release for customers. Then reverse-engineer it once you have a great example. You can cut and paste it into the prompt. Alternatively, take a previous feature release, upload it into ChatGPT, and use the following prompt:

What’s the best prompt I could have used to get this result? Rewrite it as a reusable template.

ChatGPT Output

This is super simple for illustration purposes. We’ll look at a more complex prompt later with a longer example

Write a [channel]-ready summary of a new [product_feature]. Highlight [benefit_1], [benefit_2], and include a clear call to action.

To get started first, find an example of the output you’d like—such as a sales report, blog post, whitepaper, or another reasonably complex document.

Then use a prompt that references the document—either by uploading it or linking to it online using ChatGPT which has web access capabilities.

2. Use ChatGPT Canvas to Edit and Refine

Now you’ve got the foundations for a reusable template with structured inputs. I’d also go so far as to expand this by adding an example of a product feature release and add that to your prompt. Do that using ChatGPT Canvas as described next.

With Canvas (available in ChatGPT Team and Enterprise plans, click the three dots in the ChatGPT bar to choose this) you can treat prompts like working documents:

• Break prompts into sections (Context or Role, Task, Instructions, Format, Examples)

• Test different phrasings side-by-side

• Lock in reusable blocks for future use

Canvas Prompt Example:

[Context or Role] You are a marketing strategist at a SaaS company.
[Task] Write a short LinkedIn post summarizing a new product feature.
[Format] Include a hook, benefit-driven summary, and a call to action.
[Example] “We just made onboarding 50% faster. Here’s how…”

Canvas provides a persistent workspace, letting you build and edit prompts—rather than starting from scratch each time. You can even save your prompt canvases as part of the bigger chat under ChatGPT Projects.

3. Save Final Prompts in Notion, Docs, or PromptForge

Once you’ve got a strong, tested prompt, save it. I keep all my best reusable prompts in Notion but you could use Google Docs, Evernote, or some other way to stay organized. I also use integrated prompt storage software to be able to click and paste right into my chatbots.

Notion Setup

• Create a database called “Prompt Library”

• Add the Columns: Title | Function | Prompt | Example Output | Tags | Notes

PromptForge Setup:

• Create prompt folders by department

• Add version history and variable fields

• Share with your team for collaborative prompting

Saving your best prompts turns AI from a personal tool into a repeatable team asset.

A Real Example: This Article

I used this method to create a draft of an article for the Friday edition of TheAIE using the method above. Here’s the prompt I used, note that since I had the article published online I didn’t have to include the whole article.

I started with a rough one-line prompt:
Create a prompt that replicates this result using this as the example - https://www.theaienterprise.io/p/ai-isnt-delivering

Here’s the output I got. Note that you can change the topic which I did by replacing what was in the brackets in the prompt (That’s the only change I made).

However, for illustration, the prompt below is what I got on the first shot.

Write a professional newsletter article in the style of The Artificially Intelligent Enterprise aimed at mid-level to C-level business leaders.

The topic is: [Why AI isn't delivering on its promise in the enterprise.]

Use the following structure:

Headline: Provocative, concise, 60 characters max, optimized for email subject lines.

Meta Description: No more than 145 characters.

Opening (Lede): Use a sharp insight or data point to hook the reader (e.g., slow ROI, model misuse, system complexity). Cite real-world examples or articles (e.g., McKinsey, TIME, Bloomberg, MIT Tech Review).

Body:

Break down why AI is stalling in real-world enterprise settings. Include:

Organizational challenges (e.g., change management, data silos)

Technology misalignment (e.g., proof-of-concept culture vs. production constraints)

Unrealistic expectations (e.g., AI = magic thinking)

Use 2-3 short sections with bolded subheadings for clarity.

Include stats or company examples (e.g., Gartner surveys, IBM case studies, Salesforce reports).

Solution Focus: Transition from diagnosis to strategy. Lay out 3 tactical steps for moving AI projects forward (e.g., data readiness, governance frameworks, training on AI fluency).

Close: Wrap with a concise insight or quote, offering a strategic takeaway.

Tone & Style Guidelines:

Think WSJ meets Ars Technica for enterprise.

Avoid buzzwords like "paradigm" or "game-changer."

No excessive modifiers or breathless futurism.

Avoid AI-sounding phrases. Make it feel like a senior consultant wrote it.

Keep it professional, direct, and actionable. Use tight language.

Audience: CIOs, VPs of AI/ML, Heads of Innovation, Digital Transformation leaders.

Once I used this method to get the draft I refined it using Canvas to test tone and format. I pulled in examples of a lede, the body, and the conclusion (not shown).

The whole thing took about 10 minutes once the framework was in place.

Conclusion

The difference between a so-so prompt and one that saves hours—or drives revenue—isn’t creativity. It’s structure.

Use ChatGPT to write prompts about prompts. Use ChatGPT Canvas to edit and evolve them. Use Notion or PromptForge to save the ones that work.

What you will end up doing is writing less. Prompting better. Scaling smarter.

Also another request, I want to know what you think about the newsletter. I’d love your testimonial and I’d love your critical feedback. This is so important for my newsletter and for you to get the information you want. So please take a few minutes and share your feedback. I appreciate your support.