How I draft a week of content in 30 minutes
How to run a Substack experiment so it doesn't run you.
Happy Monday!
The problem
A couple of weeks ago, I began an experiment to test Substack as a content distribution channel.
Spoiler: The results so far are pretty poor, but Substack is doing the heavy lifting for distribution, as I expected.
You’re reading this post straight “from within” this experiment. So meta…
I’m a very busy person - aside from running Postdigitalist, I write and produce content for Nada Respetable - which has received over 114K YouTube views in the past month! 🎉 - so keeping this experiment running required a pretty radical resource economy.
My situation isn’t rare. Most of the marketing teams and founders I work with have to create content in similar circumstances - that is, eager to test new stuff but spread thin.
So, how can I simplify my workflow so I hit the ground running with every post?
You know the answer: it’s AI. But I have great respect for the internet and for you. So I want to keep things, as a meme from a couple of months ago said, “very demure, very mindful”.
Principles
These are my moral constraints regarding AI:
I use AI for content production when content is downstream from product. If the content is the product, I’m personally against using AI. I don’t want anyone paying for AI slop. But, with the right fine-tuning, AI slop can be an acceptable foundation for marketing content.
AI-assisted thinking is the superior stage of “data-driven decision-making”, so it’s flawed and biased! All AI-produced content requires human oversight and insight.
Generative AI is okay, but if it’s “fed” correctly, it’s better. Examples, guidelines and long prompts are a must!
The process
I’m creating newsletter first drafts automatically, thanks to a very simple stack:
N8N
Google Sheets
Claude
Perplexity
Buzzabout
Google Docs
My brain 🧠❤
My automation
Here’s what my core automation looks like:
My trigger is a click - I click on the Execute Workflow button on N8N and we’re off to the races.
Then, N8N gets rows from a Google sheets. This is what it looks like:
My automation will filter only those posts with the “Ready” tag and send them to Claude. My Claude node is connected to a Google Docs file with in-depth guidelines and a content template.
Claude will connect my spreadsheet + my guidelines and turn them into a blog post. Then, the automation creates a Google Doc, pours the content into it and moves forward to the next post.
Before the automation
This automation only works because it’s “fed” properly. The guidelines are in-depth and carefully crafted. The newsletter ideas sheet includes:
A pain point - which I source from Perplexity + Buzzabout + my own experiences with clients
An audience - frankly, this is a pretty useless field, but I’m scared to remove it!
A case study - which I source from previously written case studies, or I write myself
A framework - an explanation of the framework, which I dictate to Claude and the tool cleans up
AI isn’t doing the thinking or the planning, it simply structures the content in a certain way. It turns untidy inputs into a “newsletter-shaped” output. This is a great use case for AI.
After the automation
The N8N automation leaves my first drafts on a Google Docs folder. There, I carefully edit them to add “the human touch”. Once the article’s live, I go to my good ol’ Google Sheets file and change the post’s status to “Published”.
Key takeaways
You know your audience “better than AI”, use generative tools to automate busywork that brings no added value, not to automate your thinking. You are irreplaceable!
You can use spreadsheets and speech-to-text to streamline content ideation.
Write long, detailed editorial standards and feed them to your generative tool of choice - it makes a world of difference.
Want to semi-automate your content production with solid editorial guidelines and a competitive content strategy? I’m taking projects for October - let’s chat.
See you on Wednesday with Framework #5.
Best,
Aaron