Save Hours on Your Newsletter With a Simple AI Workflow (No Fancy Automation Needed)
Stop waiting for creative flow. Start designing creative infrastructure.
I didn’t need a fancy automation. I needed a system.
I use ChatGPT to help with my How I AI newsletter.
(Would you expect any less from a newsletter about AI?)
I’d set it up as a Project in ChatGPT, with instructions and memory so it could keep pace with what I was doing.
That helped me feel my way through the early versions. The shifting structure, the evolving voice, the format changes.
And it worked… kind of.
I’d toss in ideas, get feedback, draft intros.
But it felt duct-taped together.
I didn’t have a process I could rely on.
No rhythm. No leverage. Just useful chaos.
Part of that was because I hadn’t defined what I actually wanted.
Over the past couple weeks, I got that clarity.
And that’s when I paused.
Not to write the next issue, but to build the machine that would write the next ten.
Here’s what I did—and how you can use the same approach to shape your own creative flow with AI.
You can’t delegate what you haven’t defined
Before you build a system, you need a signal.
You need to know what good looks like.
But too many creators jump straight to automation before they’ve wrestled with the raw material.
If I’d tried to do this a month ago, it would’ve flopped.
I hadn’t nailed my newsletter format.
I didn’t have examples to teach the AI.
I was still testing. Feeling around in the dark for something that clicked.
You have to live in the mess for a while.
Make stuff.
Miss the mark.
Notice patterns.
That’s the only way to develop the instincts you need to hand it off, whether to a person or to a prompt.
Start simple. Then improve.
Once I knew what I wanted, I was tempted to overcomplicate.
Spin up Make or n8n.
Stitch together a slick automation.
But that temptation is a trap.
First, you’ve got to learn the tools.
Then build the workflows.
Then fix what breaks.
It’s a time suck.
So I chose simple and scrappy.
I built two custom GPTs in ChatGPT:
→ One to help me find strong newsletter ideas
→ One to help me turn those ideas into drafts
Why two?
Because idea generation and writing are different muscles.
Blending them into one GPT usually gives you generic sludge. I wanted each one to be sharp at its job.
And to speed things up, I asked ChatGPT to help write its own prompts—inside the same convos where I’d been working.
That gave it real context and better footing.
Then I tested.
Round one? Decent.
Round two? Better.
Round three? Weirdly worse.
But every iteration taught me something.
I got clearer on what I needed the prompt to say and how to say it.
The hidden upside of building systems
In total, it took me about 90 minutes.
Which, yes, felt like time I couldn’t afford.
But while I was sifting through those half-baked outputs, I surfaced eight solid ideas
A full month of content at my current pace.
More importantly, I now have a tool that can give me another batch in minutes.
Then turn one into a first draft in minutes more.
And it all fits my voice, my structure, my style.
If I want to ramp up from two issues a week to three or five? Now it’s possible.
These two GPTs should take my process from 45–60 minutes per issue down to 15–20.
More output, less stress.
And it’s not finished. I can keep tweaking it. Eventually, maybe even layer in that slick automation once the foundation’s solid.
What this means for you
If you’re stuck in the loop—chasing inspiration, overthinking structure, spending hours on what should take minutes—you don’t need more motivation.
You need a system.
Not a complex one. Not a perfect one.
Just something simple and repeatable that matches how you work.
That could be a checklist.
A template.
A folder of reference drafts.
Or, like mine, a couple custom GPTs.
The real shift is this:
Stop waiting for creative flow. Start designing creative infrastructure.
Because systems aren’t the enemy of creativity.
They’re the runway.
– Nathan
P.S. What systems are you building in your creative process? Comment and let me know.