Writing LinkedIn posts from your own meeting transcripts

The idea most founders miss

Most founders think they have to come up with LinkedIn content. They sit down on Sunday, stare at a blank page, give up.

You already have the content. You generate it six hours a day. In meetings. On calls. In Slack DMs. The ideas are there. They just aren't captured.

Claude's job is to listen to what you already said and write it up as a post, in your voice.

The loop

Record a meeting. Could be a client call, a team meeting, a podcast interview. 30–60 minutes is the sweet spot.

Run it through Fathom or Granola. You get a transcript with speaker labels.

Feed Claude:

Below is a transcript of me talking about [topic]. Find 3 distinct ideas I expressed clearly enough to build a LinkedIn post around. For each, output: (a) the idea in one sentence, (b) a specific example I used or story I told, (c) a contrarian or surprising take that's implied but not stated. Don't invent anything. Quote me directly where the phrasing is strong.

What you get: three post outlines. Each one is something you actually believe, because you said it this week.

Writing in voice

The trick is the voice match. Claude, out of the box, writes generic LinkedIn slop — three short sentences, a "here's what I learned," an emoji.

To stop that, show it 3–5 of your existing good posts at the top of the prompt, then say:

Write the first post in the exact voice and structure of the examples above. Do not use phrases like "here's what I learned," "the harsh truth is," or "most people don't realize." Do not use emojis.

Two iterations and you have a post that sounds like you.

The 3-posts-a-week cadence

We aim for 3 posts a week per client on LinkedIn. Three per week, not more. More than that reads desperate.

One meeting a week can produce 3 posts, easily. That means the work is: record one meeting, drift through Claude, edit for 20 minutes, schedule. Total per week: 30 minutes of founder time.

The alternative — staring at a blank page on Sunday — takes longer and produces worse content.

Why this is better than "AI writes my posts"

It isn't "AI writes my posts." It's "AI extracts your ideas." You're still the author. Your editing is what makes it publishable. Claude's job is to make sure your best thinking doesn't stay trapped in a 60-minute Zoom call.

That distinction matters. Readers can tell when posts are AI-written (same structure, same pacing). They can't tell when posts are AI-assisted but human-owned.

What makes a post worth writing

Three things from the transcript to look for:

1. A specific story. Not "we work with B2B SaaS." A specific client, a specific number, a specific moment. If you told the story in the call, it belongs in the post.

2. A contrarian take. Something you said that most people don't. These are gold. Claude is good at spotting them — it'll surface views that are implicit but not stated out loud.

3. A framework. Three-part structures, two-by-two matrices, step sequences. If you've internalised a framework, your call transcripts are full of them. Claude can pull them out.

Posts without at least one of these three don't get engagement. Don't post them.

The biggest trap

Posting too much. Once the system works, it's tempting to go from 3 a week to 10. Don't. Your audience noticed when you posted more. The engagement dropped. Trust me — we watched this happen to 8 of 8 founders who scaled too fast. 3 per week, high quality, forever.

---

We run this for Presence clients. But the loop is simple enough you can do it yourself — most of our best-performing content starts as a 30-minute transcript.