A client of mine — Evan, who runs a members club in Austin — wanted to write a book.
He’d been thinking about it for years. He has this whole philosophy about hospitality and event design. How to arrange a room so strangers actually talk to each other. How to curate a guest list so every person feels like they belong. Stuff he’s figured out over 15 years of hosting.
But the book never got started. Too busy. Too vague. Too hard.
Then we had a working session, and about halfway through, he had a realization that stopped the conversation cold.
“If I just do 10 interviews with founders who run communities I admire… record the Zoom calls… and feed the transcripts to AI… that’s the book.”
No ghostwriter. No writing sprints. No blank page.
Just conversations. Which he was already having anyway.
He’s right. And the same principle applies to almost everything in your work.
Last year I went back through six months of my own transcripts — podcast episodes, workshop recordings, client calls, webinar replays. Fed them into Claude with one instruction: extract everything I know about AI.
Claude built a knowledge graph. Mapped out how my ideas connected to each other. Then I started asking it to write blog posts and newsletters from that knowledge base.
200 blog posts. Nearly 100 newsletters. A few days of work.
Everything was already there. I just hadn’t looked.
The Meeting Problem Nobody Talks About
When Evan and I were working through his AI setup, I asked him to prioritize. He had a long list of things he wanted to build — email automation, CRM integration, calendar management, content pipelines.
I told him: start with meeting post-processing.
Not email. Not scheduling. Meetings.
Because meetings are where everything starts. A meeting generates follow-up tasks, introduction emails, content ideas, relationship context, action items for multiple people, and sometimes entire business decisions. If you have a system that extracts all of that automatically after every meeting — you’ve solved ten problems at once.
I call this the Transcript First principle. The transcript is the source. Everything downstream builds from it.
I once built a system for a sales team where the rep would finish a call and within minutes: the CRM was updated with notes and outcome, new tasks were created in their project manager, a follow-up email was drafted and waiting for review. Thirty minutes of post-call admin reduced to two minutes of approvals.
All of it came from one transcript.
How to Apply This Right Now
Step 1: Start capturing transcripts.
If you’re doing calls and meetings without recording them, you’re letting insights evaporate. Granola works well for in-person. Fireflies or the built-in Zoom recorder for video calls. Whisperflow for voice notes on the go.
Pick one tool and start this week.
Step 2: Give AI a job before you start processing.
The way I design this for clients — and what I call Agent Design Backwards — is to start with the output you want, then work backwards through the steps an agent would need to produce it. What should a perfect meeting summary look like? What should a follow-up email look like? Define that first, then build the system to produce it.
Evan did something smart here. Before touching any tool, he drew his ideal meeting workflow on paper. Inputs, outputs, decisions. It took 20 minutes and saved him weeks of building the wrong thing.
Step 3: Treat your transcripts as a knowledge base, not a record.
The better frame: every meeting is raw material. A client call is content. A workshop recording is an SOP waiting to happen. An interview is a book chapter.
The difference between people drowning in information and people producing consistently is usually just this: they’ve built the extraction layer.
What Evan Found in His First Six Contexts
After our session, Evan went home and did something I hadn’t suggested.
Instead of writing context documents for his AI, he had a series of conversations with Claude. Let the AI ask him questions about his hospitality philosophy. About his values. About what makes a great event.
He called it “dialoging” his context instead of writing it.
By the end, he had six fully formed context documents. Including one on what he actually believes as a person — not a social media bio version, but the real thing.
He said it felt cathartic.
The Retrieval Problem
Most people think content creation is a creativity problem. Or a time problem.
It’s usually neither.
It’s a retrieval problem.
Your ideas are already out there in conversation. In calls. In workshops. In the voice notes you leave yourself driving to a meeting.
The question isn’t “what should I write about?” It’s “what did I say last week that was worth keeping?”
Build the system to capture and extract, and the blank page problem mostly goes away.
If you want help setting up a system like this — whether it’s for content, client workflows, or meeting follow-up automation — that’s exactly what we cover in my AI workshops. The next one is in Austin. Come see what it looks like when it’s working.
