My content pipeline produces blog posts, social media, and newsletters without me touching a keyboard. Here's how the automation actually works.
Last Tuesday I woke up to a finished blog draft, three LinkedIn posts queued for the week, and a newsletter ready for review. I did not write any of it. I was asleep.
This is not some magic AI tool I subscribed to. It is a content pipeline I built from agents, cron jobs, and a scheduling layer that does not care what time zone I am in.
Let me walk you through how it works.
Every piece of content in my system goes through the same five steps.
Step 1: Topic selection. A cron job checks the content calendar every Monday morning. The calendar is a structured document with topics, target dates, target platforms, and content pillars. The job pulls the week’s topics and creates task briefs for each one.
Step 2: Research. A research agent takes each topic and gathers supporting material. For a blog post, that means relevant data points, examples, and any of my past content on the subject. For a LinkedIn post, it pulls recent industry conversations worth referencing. The research output is a structured brief: key points, supporting evidence, recommended angle.
Step 3: Drafting. A writing agent takes the research brief and produces a draft. It follows a voice guide that defines tone, sentence patterns, formatting rules, and words to avoid. The draft is not a final product. It is a working version that gets the ideas down in the right structure with the right voice.
Step 4: Review queue. The draft goes into a review folder. I check this folder once a day, usually in the morning with coffee. Some drafts need light editing. Some need a heavier pass. Some are good to go. The human review step is non-negotiable. I will never publish content I have not read.
Step 5: Scheduling. Approved content gets pushed to the scheduling tool for publication at the right time. Blog posts go to the CMS. LinkedIn posts go through Postiz. Newsletter content goes into the email platform.
Start to finish, the pipeline produces three to five pieces of content per week with roughly 30 minutes of my time per day for review and editing.
The exact architecture, memory layers, and delegation patterns I use to run 50 agents across two businesses.
Get the AI Agent Blueprint →Content is uniquely suited to automation because it is repetitive, structured, and high-volume.
Every blog post follows a pattern. Introduction, key sections, examples, conclusion, call to action. The structure is predictable even if the content is different each time. Agents handle predictable structure well.
Every LinkedIn post follows a different pattern but still a pattern. Hook, body, takeaway. The format is fixed. Only the message changes.
Newsletters are the same. Subject line, intro, sections, sign-off. Templated structure with variable content.
Where automation fails is originality. An agent cannot have a conversation with a client and turn that into a story. It cannot notice something in the news and connect it to a business insight I have not already documented. It cannot write from personal experience it has not been given.
That is why the research step matters so much. The quality of the output directly depends on the quality of the input. Feed the writing agent a thin brief and you get generic content. Feed it a detailed brief with specific examples, data points, and a clear angle, and you get something that sounds like a human wrote it. Because, in a meaningful way, one did. My thinking, my examples, my perspective, assembled by an agent.
Generic AI content sounds like generic AI content. You have read it. The vague claims. The predictable structure. The vocabulary that no human actually uses in conversation.
My voice guide is a document that defines exactly how content should sound. It specifies things like:
Sentence length variation. Short punch sentences mixed with medium explanation sentences and occasional longer builds. Monotonous sentence length is the fastest way to sound artificial.
Words to avoid. There is an explicit ban list. Words like “delve” and “utilize” and “comprehensive” and “robust” that nobody uses in real conversation but AI models love. If those words appear in a draft, the agent rewrites the sentence.
Formatting rules. Short paragraphs. Bold text for emphasis on key terms. Bullet points for lists, not run-on sentences. Headers that describe what the section delivers, not clever wordplay.
Perspective and framing. First person. Direct address. Concrete examples over abstract principles. Real numbers over vague claims.
The voice guide is roughly two pages long. Every content agent has it loaded as part of its context. The result is not perfect, but it is close enough that 10 minutes of editing produces something I am comfortable publishing under my name.
Without a content calendar, the pipeline has nothing to do.
I plan content monthly. Four content pillars: Architecture, Automation, Building in Public, and Strategy. Each pillar gets roughly equal representation across the month.
The calendar is a table with columns for date, title, pillar, target platform, and status. The cron job reads this table, creates tasks for anything with a status of “scheduled” and a date within the current week, and the pipeline takes over.
Planning the calendar itself takes about an hour per month. I batch it. One sitting, map out the month, done. The rest of the month, the system executes.
Running the content pipeline costs roughly $40 to $60 per month in API calls. That covers research, drafting, and any revisions.
For context, a freelance writer producing the same volume (three to five pieces per week of equivalent quality) would cost $2,000 to $4,000 per month.
The tradeoff is my time spent on setup and maintenance. The pipeline took about a week to build initially. Ongoing maintenance is maybe an hour per month, mostly tweaking the voice guide or adjusting the calendar format.
At these numbers, the pipeline paid for itself in the first week.
You do not need my full system. Here is the minimum viable version.
Create a content calendar. Even a simple spreadsheet with dates and topics.
Set up one writing agent with a voice guide. Use any LLM API. Give it a system prompt that defines your voice, your formatting preferences, and the output structure.
Create a review process. Even if it is just a shared folder you check once a day.
Schedule manually at first. Copy the approved content to your platforms. Automation here is a convenience, not a necessity.
That is it. Four components. You can have this running in an afternoon. The sophistication (research agents, automated scheduling, quality checks) comes later as you see what works and what needs improvement.
The blueprint includes the full pipeline architecture, including the voice guide template and calendar format.
Content compounds. The sooner you start producing consistently, the sooner the results show up. A pipeline makes consistency automatic.
Subscribe to The Signal
One issue per week. The biggest AI build of the week, the top five videos worth your time, and an operator's read on what actually matters. For builders who ship, not dreamers who scroll.
No spam. Unsubscribe anytime. Just The Signal.