Creating a Content Community Lead with AI: What I Built on Saturday

We all do it. A friend launches something. A company you love is hiring. Someone in your network needs a signal boost. You want to help, but writing the post feels like a chore. You open LinkedIn, stare at the box, type three words, delete them, and close the tab.

Multiply that across a year and you've broken hundreds of small promises to people you actually care about. Not because you're a bad friend. Because the friction is too high.

This is the exact kind of task AI should be eating. It's low-stakes, it's repetitive, and the thing that makes it hard isn't the writing - it's the context switching. You have to remember why you care, who your audience is, what your voice sounds like, and what's already been said. By the time you've loaded all that into your head, the moment has passed.

The trick is to stop loading it into your head and start loading it into a system that already has it.

The single biggest unlock for me over the last year has been building what I call my Batko Brain. It's a GitHub repo that holds everything about me. Public stuff: blog posts, podcast transcripts, talks I've given, articles I've been quoted in. Private stuff: notes from coaching sessions, personal frameworks, decision logs, retros from things I've shipped.

I structured it cleanly so an AI agent can actually read it. Folders by topic. Markdown everywhere. Timestamps on the things that age. A `LinkedIn-style-guide.md` that captures the way I write hooks, how I structure posts, the cadence of short and long sentences I tend to use.

When I want to write a LinkedIn post now, I don't prompt an AI to be me. I point it at the brain and ask it to draw from me. Those are very different things. One produces a generic AI impression of a personal brand. The other produces something I'd actually publish.

The prompt is the smallest part of the work. The brain behind it is doing 90% of the lifting.

If you're starting from scratch, the minimum viable brain is three files:

• A bio with the things you want to be known for
• A style guide with two or three of your best posts pasted in as examples
• A running list of the experiences and stories you draw on most often

That alone will move your AI output from forgettable to distinctly yours.

Here's exactly what I did during the Build Hour, step by step. No magic.

First, I gave the agent the context. The Startmate role description, the fact that I have a personal connection to the company, and a note that this is a job I'd genuinely want my network to see.

Second, I pointed it at two things in my brain: my LinkedIn style guide and the section of the brain that has notes from my own time around the Startmate ecosystem. Old conversations, lessons I'd written down, people I'd worked with there. The agent now has flavour, not just facts.

Third, I asked for a specific output. Not "write a LinkedIn post". I asked for five different hooks and four full post drafts, each taking a different angle. One straight shoutout. One self-deprecating nostalgia angle that tags the people I worked with. One "taste is the most underrated skill" angle. One "here's the kind of person who'd thrive in this role" angle.

Fifteen seconds later I had four drafts in front of me. Not all of them were keepers. The first one was solid but a bit stiff. The taste angle was genuinely good and I'd never have written it from a cold start. The nostalgia one made me laugh out loud. That's the bit that matters. When AI surfaces an angle you wouldn't have found alone, it stops being a writing tool and starts being a thinking partner.

I picked the one I liked best, edited maybe 20% of it, and scheduled it. Total time from "I should post about this" to "posted" was under four minutes.

The mistake I see everywhere is people treating AI as a content generator instead of a content amplifier. They open ChatGPT, type "write me a LinkedIn post about leadership", and get back a piece of bland sludge that could have been written by anyone, about anyone, for anyone. Then they conclude AI writing is rubbish.

The AI isn't the problem. The brief is.

If you want output that sounds like you, the input needs to be you. That means:

• A clear voice reference (your best past posts, ideally five or six of them)
• Specific personal experience the agent can pull from
• A sharp angle, not a topic
• A specific format, not "a post"

Most people skip all four and then blame the model. The frustrating part is that doing the prep work once unlocks every future post forever. It's a one-time cost that pays out for years. I genuinely think that within 12 months, not having a personal context brain is going to be the new not having a website. It will look strange. It will quietly hold you back.

And if you're doing it for a business, the same logic holds, just bigger. A company brain with positioning, customer stories, internal language, and product detail will make every piece of marketing content the team produces dramatically better.

I'm not going to paste the final post here because by the time you read this it'll be live on my LinkedIn anyway. But I'll tell you the angle I picked and why.

The winning angle was taste. The line that came out of the agent was something like "the Content Community Lead role at Startmate is one of the few jobs left where taste is the actual skill being hired for". And that landed for me because it's true. Most marketing roles in 2026 are being pulled apart by automation. The bit that's becoming more valuable, not less, is the human judgment about what to make, who to make it for, and why it matters. Taste is the moat.

That's not a thing I would have written from a blank box. I'd have defaulted to "hey, my friends at Startmate are hiring, please apply". Functional, forgettable. Instead I got something with a point of view. Something that actually says something about how I see the world.

This is the part of AI-assisted writing that I think gets undersold. People talk about speed. The real win is angle discovery. Having a thinking partner that can read your whole context and surface the framing you would have missed.

If you take one thing from this post, let it be this: don't use AI to replace your voice. Use it to scale your voice across the things you never get around to.

Here's the smallest possible version of the workflow you can build this week:

• Create a folder somewhere. Call it `my-brain`. It can be Notion, GitHub, Google Drive, Obsidian, whatever you'll actually open.
• Add a `bio.md` with three paragraphs about who you are and what you care about.
• Add a `style-guide.md` with five of your best past posts pasted in.
• Add a `stories.md` with a running list of the experiences you draw on most often. Update it any time something happens.
• The next time you need to write a post, paste the role or context into your AI of choice, point it at those three files, and ask for five hooks and three full drafts in different angles. Pick one. Ship it.

That's the entire system. You can build it in an afternoon. It'll save you hundreds of hours over the next year and your output will get noticeably more you, not less.

The Build Hour livestream is going to keep happening every Friday at 3pm. I'm also bringing it into different offices for live Q&A sessions. Notion is the first one locked in. If you'd like me to do one from your office, just reach out. The whole point is that this stuff is practical and the only barrier most people face is inspiration. You need to see someone else doing it before you realise it's possible. That's the moment everything changes.