Million Dollar YearGet the free app

How to build an AI content machine with n8n, Airtable and Slack

By Max Warnault · 24 July 2025 · 10 min read

In short

An AI content machine is a set of n8n automations tied to an Airtable database and a Slack channel. It collects ideas from competitors every week and from posts you share. An AI agent with your brand details turns each idea into carousels, image posts, short form scripts and long form YouTube outlines or scripts. A publishing automation then posts the queue to every platform at noon each day. You review the drafts and record the videos.

Making content for Instagram, X, Threads, TikTok, Facebook, LinkedIn and YouTube every day eats a lot of time. Most of it goes on writing each post from scratch and publishing it by hand.

I built a system that does most of that work for me. Ideas come in on their own. The AI writes each piece in my brand voice. Every post gets published on every platform without me touching it. My content work is down to a few hours a week. When I recorded this video I had 42 posts in the queue, so about 42 days of content ready to go: images, carousels, reels and long form YouTube videos. Here is how it works, from the front end to the automations behind it.

What the AI content machine does

Everything lives in an Airtable database. It has a few main parts:

  • An ideas backlog, which includes my competitors' content, added automatically
  • A brand section that tells the AI who I am and who I help
  • The content itself: carousels, image posts, short form scripts, long form outlines and scripts
  • A publishing queue that posts to every platform each day

I talk to the system through a Slack channel I call content marketing. The automations run in n8n, an automation platform that lets you automate pretty much anything you want.

The brand section that makes every post sound like you

The AI can only write like me if it knows me. So the brand section holds:

  • The brand name
  • The presenter: who I am and my story (I kept this short)
  • The prospects I help
  • The problem I solve
  • The promise I make
  • The process I use to solve that problem
  • The products I sell
  • The pitches: the calls to action I want on my posts, so I can send people to different lead magnets or products depending on the topic
  • The content strategy
  • The brand identity, brand pillars and brand personality

Once that is filled in, everything the system writes is branded to me. That includes content it rebuilds from someone else's idea.

Where the content ideas come from

I don't reinvent the wheel. Ideas come in four ways.

  1. Competitor content, every week. I add a competitor's name, their Instagram and their YouTube to the database. Each week the system pulls in their content from the past seven days and adds it to my ideas list. If I like a post, I can recreate it for my brand, my audience and the problem I solve.
  2. Posts I share from my phone. When I scroll Instagram or YouTube and see something good, I share it into the Slack channel and it lands in Airtable as an idea. One example was a long form video from another creator about why you need systems instead of goals. That fits my brand, so I saved it to make my own version.
  3. YouTube research. I can ask in Slack, for example, "please find YouTube content on n8n MCP servers," and it brings back the videos.
  4. Original ideas. I type or record what I want in Slack and the system builds a new piece of content from it.

How a carousel gets made, step by step

To show it live, I asked for a funny carousel. The idea: I've never shared this publicly, but I live a double life. Supermax by day, building AI systems to scale businesses. Superman by night, saving cats and grandmas.

  1. I typed the idea into Slack and hit enter. The system confirmed it had captured the idea and added it to Airtable with a full slide plan: a minimalist design, setup text for the confession, and the text and image idea for each slide.
  2. I asked it to create the carousel text. It turned the plan into the full slide-by-slide text and saved it as a child piece of content linked to the original idea.
  3. I quality checked the text and edited the lines I wanted to change, for example which line opens slide one and which goes on slide two.
  4. For custom pictures of myself I use Flux. I set the ratio (the carousel needs 4x5) and wrote prompts like "Max, bald, with glasses, disguised as Clark Kent working on his computer in a professional environment." I say I'm bald in the prompt because otherwise it sometimes gives me hair. When an image came out weird I raised the number of steps or made the prompt simpler. In general, the simpler the prompt, the better the image.
  5. I ran the image automation. The slide images, in a Twitter post style, appeared in Airtable after a few seconds. I don't open Canva to build them.
  6. The last step is manual. In Canva I would swap slides three and four for the two Flux images, add the text, upload the finished carousel to Airtable and put it in the queue. From there it gets posted automatically.

I chose no call to action on that post. The pitches in the brand section let me add one whenever a post needs it.

The back end: how the system works in n8n

Here is what runs behind the Airtable and Slack front end.

The main trigger

Every request goes through one main automation. A Slack trigger watches the channel. Any time I post there, it checks whether I sent text or audio. If it's an audio file, it downloads it and transcribes it. That means I can send a voice message like "create a piece of content on why every business owner needs an AI system to manage their content creation" and it works.

The message then goes to an AI agent that has access to an MCP server. The same request can also start from Airtable. When the agent finishes, it replies with what it did, for example "I've successfully created the carousel." I click the child content to see the result.

The weekly competitor spy

A separate automation runs on a schedule, every Friday at 6 a.m. This one was a beast to build.

  1. It gets the list of competitors I want to watch.
  2. It pulls their YouTube videos and their Instagram posts.
  3. If an Instagram post is a video, it adds the video and transcribes it. If it's an image, it reads the text on the image and transcribes that.
  4. When a transcript is ready, it comes back through a webhook. The system then generates a title and writes the text onto the idea in Airtable.

From any of these ideas I can create a carousel or a short form video with one request. Each new piece is saved as a child of the original idea.

The MCP server and its tools

Whenever the AI agent is called, it can use every tool on the MCP server:

  • Import an Instagram post
  • Import a YouTube video
  • Research YouTube
  • Manage content in Airtable: get the record ID, get the content, list content, delete content, add new content and update content

For example, I asked "please show me all the content I need to record." The agent went to the MCP server and called the list tool. It wrote the Airtable formula that shows only the content waiting to be recorded, then gave me the list of video titles.

New content

When I send a new idea, the system generates a title. It searches Perplexity to expand on what I said, then creates the Airtable record.

Importing Instagram and YouTube, and handling long transcripts

For Instagram, it pulls the reel and transcribes it, or transcribes the text on an image, then adds a title from the description and updates the record.

For YouTube, it gets the video link and the transcript. Sometimes the transcript comes back right away and sometimes it doesn't. When it doesn't, a second automation uses another tool to get it.

Long transcripts were the hardest part. Airtable has a character limit, and a three-hour podcast transcript won't fit. So when a transcript is too long, the system puts it in a Google Drive file, sets the permission so anyone can read it and saves the link in Airtable. I also had to edit every other automation to check for that link and pull the transcript from the file. This way I always have the transcript.

Updating content with a custom prompt

I can change any piece of content by typing a custom prompt, for example "turn this into a 10 slide carousel instead of two." The update automation gets the record, makes the change, sends it back to Airtable and tells me when it's done.

Carousel text and carousel images

The carousel text automation checks whether the transcript is in a file. If it is, it pulls the text from the file. Then it uses Perplexity to expand on it and generates the carousel and its title. You quality check it and make changes before any image is made.

The carousel images use another API. The text has a break character between slides. The automation splits the text on that break, loops over each item and creates one image per slide. If you want a background or a different style, you just update the automation.

Short form scripts

Short form can come from a YouTube video, another short form video or a carousel. For a long YouTube video, one generic reel would waste most of it. A creator might share many different ideas in one video. So the automation analyses the transcript and pulls out the separate reel ideas. For each one it writes a new reel script, title and caption and uploads them to Airtable. One long form video becomes many short form scripts. If the source isn't from YouTube, it writes one reel with its title and caption.

Long form outline and long form script

The outline automation summarises the text and decides whether it needs more research. Content from YouTube usually doesn't. It only needs rewriting around what I want to say. Then it creates the outline and uploads it to Airtable. Some of my clients only want bullet points when they record YouTube videos, and this gives them that.

The script automation turns the outline into a full script that's ready for a teleprompter. You read it word by word and upload the video to YouTube.

Publishing every day at noon

The last automation publishes content every day at 12 p.m. I post short form videos, images and carousels each week, plus one long form YouTube video. Each type can only go to certain platforms. You can't post a carousel on TikTok, for example. So I set up each branch of the automation for its content type and the platforms that accept it.

Short form goes to Instagram, YouTube, TikTok, a personal Facebook page, a business Facebook page, Threads, Twitter, personal LinkedIn and a business LinkedIn page. Images and carousels go to the platforms that accept them. The long form video goes to YouTube every week.

Why this lets one person run all the content

I'm the only person running all of this. I don't need a team, because the system manages all my content in one place. I built it once, and now it works every day, posting on every platform. Once or twice a month I create the images, carousels, short form videos and long form YouTube videos, and they get uploaded everywhere.

This is the kind of code leverage we build at Million Dollar Year: a system that keeps working while you don't, so you can grow without hiring.

Key takeaways

  • Fill in a brand section (story, prospects, problem, promise, process, products, pitches and personality) so every AI draft sounds like you.
  • Add your competitors' Instagram and YouTube accounts and pull their past week of content into your ideas list on a weekly schedule.
  • Use a Slack channel as the front door so you can send ideas as text, shared posts or voice messages.
  • Have the AI write carousel text first, check it yourself, and only then generate the slide images.
  • Split long YouTube transcripts into separate reel ideas so one video turns into many short form scripts.
  • Store transcripts that are too long for Airtable in a Google Drive file and save the link on the record.

Questions and answers

What tools do I need to build an AI content machine?

My system runs on n8n for the automations, Airtable as the database and a Slack channel as the place I send requests. It also uses Perplexity to expand on ideas, Flux for custom images of myself and Google Drive to store transcripts that are too long. An AI agent connects to an MCP server that holds all the content tools.

How do I turn competitor content into my own posts?

I list my competitors with their Instagram and YouTube, and every Friday at 6 a.m. an automation pulls their content from the past week. It transcribes videos and reads the text on images, then adds each one to my ideas list. From any idea I can ask for a carousel or short form video rebuilt for my brand, my audience and the problem I solve.

Can I create content by sending a voice message?

Yes. The Slack trigger checks whether I posted audio. If I did, it downloads and transcribes the file and sends the text to the AI agent, which creates the content in Airtable and replies with what it did.

How do I post the same content to every social media platform automatically?

A publishing automation in n8n posts the queue every day at 12 p.m. Each content type has its own branch, because some types can't go everywhere. A carousel can't be posted on TikTok, for example. Short form goes to Instagram, YouTube, TikTok, Facebook, Threads, Twitter and LinkedIn, and the long form video goes to YouTube every week.

Do I still have to do anything by hand?

Yes, a little. I check the text before the images are made, swap in any custom images in Canva and record the short form and long form videos from the scripts. I do that once or twice a month, and the system handles the posting.

For 6-figure coaches and agencies

Build a million dollar business, solo, with systems and AI.

Start with the free app: your year planned down to today, your numbers in one place, and exactly what your goal takes.

Get the free app

Free · No card · On the web and on your iPhone

Keep reading