If you've found yourself a little overwhelmed using AI to create content...
Maybe you don't know what to say to get something useful and consistent out of it—or even where it should fit best into your workflow. The good news is, there's a tactical answer to this problem. One that gets you interacting with chat-based AIs more effectively overall.
This isn't about the prompts. It's about narrowing the focus of the AI, and redirecting it with specificity during the creation process.
What you're about to read is one tactical execution of this principle.
There are many ways to collaborate with AI in the creative process—too many to cover in one article. Use this as your starting point.
How to use what you've already done to accelerate your work
Quick context: AI creates text by predicting what words go in what order based on what it has already seen.
So if you hop right into GPT and ask it to write an email, it's going to pull from it’s entire language model and give you something... average.
To get it to write what you need it to write, you have to narrow its focus.
Here’s the tactic:
(Using email as the example, but this works for any kind of content.)
Focus on repeating one thing. If you want to create a consistent stream of weekly broadcast emails, start by identifying your five best-performing ones.
Think about the outcome you want to repeat. If it's opens, pick your most opened emails. If it's clicks, use your most clicked. If it's revenue, focus emails which resulted in sales.
The key is: focus on one type of asset, and one type of result. (You could be creating series, use series as references. You could be creating landing pages, etc)
Then: {NOTE — Using ChatGPT as the example, you can repeat this with most of the chat based AI programs}
Step 1: Open a new project in ChatGPT (other chat based AIs have a similar feature, where the reference is narrowed to the “project” and project files). Add your five selected emails as project files. Prompt it with something like:
"I want you to analyze the 5 emails I've added to the project files. These emails all had the highest open rates on our list. Identify the patterns across subject matter, writing style, big ideas, hooks, subject lines, preheaders - everything they have in common. The goal is to repeat these results."
ChatGPT will give you its analysis.
Step 2: Compare its analysis to your own.
This part matters. Don't treat AI like a teammate you can trust to think for you. Critically review what it gives you. Tell it what you agree with, what you don't—and why. If it missed something you think should be there, tell it what that is and why.
It will take all that feedback and adjust.
Step 3: Once you're happy with the analysis, ask ChatGPT to compile everything into a single framework document it can reference when writing future emails.
Double-check it for accuracy. Then add it to the project files.
Creating with The Framework
Now when you write with the emails, we can start with the framework, but that’s only the beginning.
Always narrow the AI’s focus. Whenever possible, feed it your insight, your experience, your judgment.
The more specific your inputs the more specific your outputs.
And the more consistently you do this inside a single project, the better it gets. GPT will start producing assets that are closer and closer to what you want.
It will grow with you.
This isn’t about treating AI like an output machine—or even a person. (Though speaking to it like a person is wildly effective... but that’s another article.)
Think of it like the learning loop.
The AI Loop
Say you’ve got a new promo email to write.
You open up ChatGPT, and inside your project you already have:
- The framework you created
- A few winning emails as references
- A rough outline or thought dump on your new idea (goal, topic, direction, etc.)
Then you prompt it:
"Using our email framework, write a new email based on my thoughts below."
The first version might be close—or way off.
Most people get stuck here. They start fiddling with prompts, hoping it gets better. But the truth is, you probably already know exactly what you're looking for in your head.
The more specific the outcome you want, the more you'll need to be involved.
So treat that first output like a draft.
Give clear feedback. Rewrite sections (or even just add your notes/reactions to each section). Make notes on pacing, tone, word choice, formatting - whatever it needs.
Feed those changes back in. Ask for a revision.
Keep looping.
Even with the framework in place, the more you actively direct the model on nuance - tone, tension, formatting, idea structure - the more accurate it will get over time.
You're not just getting a better email.
You're training the model to work like you.
Not "write for me."
But "write with me."
In addition to the dynamic of the AI getting closer to the mark each time you create, this process also allows you to quickly create content useful using the unique ideas in your head without draining you of your creative energy.
(But, that's also another topic for another time)
Why This Works
Most of the content you create is some kind of repeatable asset.
This is the same principle behind the "Rear View Mirror1" framework. Even when you write from scratch, you're probably pulling from what’s worked before—things your audience responds to, rhythms that resonate, ideas that convert.
If you can figure out why something worked once, you can make it work again.
For email writing - let’s say it’s a weekly broadcast - you go into yourself knowing roughly the style, structure and goal, and you enter that with a specific idea. Maybe it’s something you came up with, maybe it’s from a video. You can write your thoughts down for the AI to use, or you can feed it a transcript where the ideas are talked about. There are tons of different little ways to use this tool effectively.
Most people essentially end up use ChatGPT to guess.
I use it to rapidly reverse-engineer.
In many ways it's like having an analyst, a strategist, and a writer all in one ...
But in reality it's an external extension of me.
In the long run it works because it’s not about replacing creation.
It’s about building a repeatable system that sharpens over time.
🎥 Want to see this in action?
Do you want a short Loom video where I walk through how I used this exact process to create this article?
Comment below and I’ll DM it to you directly.
Be Useful. Be Present. Love the Journey.
Joseph Robertson, CMO Man Bites Dog
Ready to Step Into The Arena?
Ready to engage the field? Man Bites Dog paid subscribers have comment access unlocked below. (They’re also sent a killer welcome package in the mail with all kinds of opportunities that are not available in digital format)
Here are some other options:
Get on the waitlist to join the Arena: engagethefield.com
Check out the Engaging The Field Handbook
Grab your your own copy of the R3 system (it’s a book and it’s not cheap)
Love the intention and integrity of this process, Joseph. Augmented Intelligence > Artificial Intelligence (for sure).