I can already hear some of you saying, "Kyle, what have you done to me? I set up those Zapier feeds like you suggested, and now my spreadsheet is exploding with content. This is worse than before!"
First off—I apologise! I just opened up a firehose of information and pointed it straight at your Google Sheet. That wasn't very considerate, was it?
But hold on before you close this tab in frustration. The database we built yesterday isn't meant to be your final destination for content review. It's just a holding tank—a repository where everything gets collected before the real magic happens.
We've built the plumbing to bring water into the house, but we haven't installed the filters yet. That's what we're doing today - adding a crucial filtering layer that transforms an overwhelming flood of content into a manageable, curated stream of high-quality candidates for your newsletter.
Sounds good? Sorry we had to do set up the firehose first! But we need the holding tank first!
By the end of today's session, you'll have an AI system that reads everything for you, summarises it, rates its relevance, and presents you with only the most promising items. Your morning review will go from overwhelming to effortless.
Let's get started:
Even with just 5-10 sources feeding into your content database, you could easily end up with 30-50 new items each day. Reading all of that would defeat the purpose of our efficient atomic newsletter system…
This is where AI becomes our essential ally. We're going to build a filtering layer that processes this information flood and distills it down to only the most relevant and valuable content for your specific newsletter.
First, let's make sure we're properly setting up new content with the correct status. In Let’s adjust the Zapier action that adds content to your Google Sheet (that we made in Part 2).
You need to explicitly map "New" to the Status column:
In the Google Sheets action of your Zapier workflow:
We’ll use this to trigger actions shortly.
If you’re new to AI Automation and Zapier, check out our AI Automation Accelerator.
Before we dive into the technical setup, we need to clearly define what makes content worthy of inclusion in your newsletter. This will form the basis of our instructions to the AI.
And it depends on what type of newsletter you are building.
Remember the newsletter type you chose in Part 2? Your content criteria should align with that:
Feel free to use these or write up your own criteria! We’ll use them in the prompt below.
Now, let's add an AI layer to the content collection system we built yesterday. Here's the basic workflow we're creating:
For this setup, we'll use ChatGPT (either via API if you're technical, or manually if you're not). If you prefer Claude or another model, the same principles apply.
If you're not comfortable with APIs or coding, you can still leverage AI effectively with a manual approach:
You are an expert content curator for a B2B newsletter focusing on [YOUR INDUSTRY/NEWSLETTER TYPE].
I'll provide you with a batch of content items. For each item, I need you to:
1. Create a concise 2-3 sentence summary capturing the key insights
2. Rate the item from 1-5 stars based on these criteria:
[INSERT CRITERIA FOR YOUR NEWSLETTER TYPE FROM ABOVE]
3. Explain your rating in 1-2 sentences
4. Suggest a brief insight I could add if I featured this in my newsletter
Here are the content items to analyse:
ITEM 1:
Title: [TITLE]
Source: [SOURCE]
Content: [PASTE FULL CONTENT]
ITEM 2:
Title: [TITLE]
Source: [SOURCE]
Content: [PASTE FULL CONTENT]
This manual approach takes about 10-15 minutes each morning but saves considerably more time in review and selection.
I actually suggest starting with this to test out the quality of the output. Then when you are happy you can build it into the automatic flow.
If you're comfortable with Zapier, we can create a more automated version. Here’s the basic flow:
The extraction can be done using Zapier's Formatter tool with text parsing functions to pull out the relevant sections from the AI's response.
This approach requires more technical setup but runs automatically each morning with no additional input. Once built it’ll run everything for you.
Let's implement this workflow using Zapier. We'll set up two main Zaps:
This setup automatically identifies new content, sends it for AI processing, and updates your database with structured, analysed content.
After implementing this system, your content database will transform from a chaotic collection of raw material into a structured, analysed resource.
Now you have a database where:
With this filtered database, you've dramatically reduced the noise and highlighted the signal. Instead of wading through dozens of full articles, you can now scan this structured table and quickly identify the most valuable content items.
At this point, you have a solid content filtering system:
BUT it’s still in a database which is a bit….busy!
Tomorrow in Part 4, we'll build on this foundation to create: