AI Portfolio
Content Studio Agent
One idea in, every channel out, one human approving
The Problem
Content stops the moment client work starts
Content is the first thing that slips when I get busy.
A client project lands and for a week or two I keep posting. Then I'm fully in it, and the channels go quiet. Months later the project wraps, I look up, and the blog hasn't moved, LinkedIn has gone cold, and the YouTube channel I keep saying I'll start still isn't a thing.
The smaller problem, and the one that actually bugs me, is that one idea usually becomes one post. I think something through, put it on LinkedIn, done. It could have been a blog article and a video too. I just never get that far.
So it was never really about whether AI can write a post. It's that one idea should travel, and the content shouldn't stall every time I don't have a free evening.
The Pipeline
One idea in, three channels out
I set it up like a real content team, not a prompt I paste into.
It starts with one raw idea: a thought, a link, a voice memo, half a paragraph. The content-studio agent reads it and decides which channels it should live on. Not everything deserves a blog post. Some things are a LinkedIn post and nothing more.
From there it splits. Content Refinery writes the blog article and the LinkedIn post in my voice. YouTube Script handles the video side. Everything lands in one Notion Content Calendar, which is the single source of truth for what exists and what state it's in.
Then it waits for me.
There are two review gates, and I own both. Nothing publishes silently. I approve the angle first, then I approve the finished pieces. When I say go, each piece routes to where it belongs: the blog publishes as a pull request with a cover image, and LinkedIn goes out through Buffer.
The YouTube leg is wired but not running yet. That's the honest status, and there's more on that below.
Under the hood
The agent is thin. The context is everything.
The agent itself doesn't hold much. What makes the output sound like me is the layer underneath.
Before it drafts anything, the agent reads the Second Brain in Notion: my brand voice, who I'm actually writing for, the content pillars, and the post structures that have worked before. That's why the writing doesn't drift. Every piece starts from the same understanding instead of me re-explaining myself every session.
Underneath the agent sits a library of more than a hundred skills it can call: the specific recipes for writing a post, outlining an article, scripting a video.
And the loop closes itself. Weekly Pulse pulls my LinkedIn and GA4 numbers every week and writes them back into the Second Brain, so the next batch of content knows what landed and what didn't. The system gets a little less wrong over time, without me running the analysis by hand.
Where I Am Now
What's live, and what's still me
Honest status, because pretending on my own site would be silly.
Live and running
- **Blog leg.** One idea becomes a full article in my voice, published as a pull request with a generated cover image.
- **LinkedIn leg.** Posts drafted and scheduled through Buffer, on my approval.
- **Notion Content Calendar.** The single source of truth, one row per piece, status from idea to published.
- **Second Brain + Weekly Pulse.** Context in, performance data back, every week.
Not there yet
- **YouTube leg** is wired but I haven't run it end to end.
- **The full pipeline in one pass**, one idea cleanly through all three channels, I've tested in pieces, not as one smooth run.
The funny part is where the bottleneck sits. The agents are fast and the writing is good. The thing that slows the whole system down is me: approving, deciding, finding the evening to look. I keep saying the bottleneck is never the building, it's the deciding. Turns out that applies to my own setup too.
Further reading
Most Companies Use AI to Shrink. The Smart Ones Use It to Grow
Most companies are using AI to cut headcount. That's the lazy half of the equation. Real AI transformation is a business redesign, not a tech project.
The Bottleneck Isn't Building. It's Deciding.
AI made execution fast. But approval still runs on slide decks and steering groups. The bottleneck in product development isn't technical — it's organizational.
What Most Product Teams Don't See About Business Systems
Most product teams see data, tech, and product. But businesses operate in thirteen interconnected layers. Here's how to see the whole system.
Get the next article in your inbox
Short, practical notes on strategy, operating models, and what it takes to turn intent into something that actually works.