AI can generate content quickly, at low cost. It's a good starting point. But how do you reach quality and, above all, automate it to produce quantity?
That's what I show you in this section.
I've been writing on Substack for more than a year (over 60 editions). I share concrete feedback on using AI in business. This section is hard to automate. However, I have a section further down on the latest news, and that block is 100% automatable and doable in any market. That's what I show you here.
Hopefully this gives you ideas for other text formats (blog article, social media posts, e-book, and more)
Reminder: my newsletter is not automated, and this section is the result of hands-on experience automating a news newsletter for a client company.
The goal
Set up a complete automation to produce and send a weekly newsletter on a sector's news.
The workflow will rely on 3 tools:
- Tavily → search and extraction of raw content (articles, images, sources). There are alternatives such as Linkup or Exa : tools able to search and extract web content from a text request.
- OpenAI (or Claude) → content generation and validation
- Brevo → sending the newsletter with human review before release
Why run a newsletter?
It's a very good retention channel where you can both share value with your market and relay the offers you provide, regularly.
It's also a chance to become a reference in your market, share your convictions, and build a mini-media. This retention channel can turn into an acquisition channel depending on how popular your content becomes.
This sharing has helped me attract a little over 14,000 readers today (thank you!). It has become a pillar of my business, so I encourage you to launch your own newsletter. AI can help you!
The workflow structure
The idea is to combine Tavily for news search with an AI model like OpenAI that can identify the relevant news within a set of search results.
Once the news selection is done, we gather as much content as possible to share with a model and generate content in a precise format.
The more context you share with a model, the better your chances of producing complete, high-quality content.
We group the different news items to generate a coherent, final section before integrating it into a newsletter template.
So we take this diagram and build it in a workflow automation tool like n8n. Here is the result:
A closer look at the steps
We use Tavily to retrieve relevant URLs from a simple query:
Once that's done, the idea is to use AI to prioritize the most relevant results against a newsletter history, the target audience, and so on (per your instructions)
For the advanced research work on each news item, we retrieve the full content of the articles and even the image URLs (which we will integrate into the newsletter)
We add everything to an emailing tool to use a template and fill in the fields. For example, in this template there is a first “FOCUS ON” section and “NEWS” sections. The last “OFFER” section lets you highlight one of your company's offers.
At the end of the workflow, we get the template filled with the different sections. Here's an example:
To go further, I invite you to watch this video which explains the whole process step by step and even shares the complete workflow with you.
Why not automate?
I'm talking to you about automation even though this newsletter isn't automated. Mainly because the format I aim to offer here is centered on concrete, hands-on feedback (and because I enjoy writing it).
In some situations, automating a newsletter isn't suitable at all (especially if you choose a tool like Substack that has no API).
In others, ones that can relay news content or snack content, it's highly relevant.
It's up to you to decide whether it can be for you or your company :)
