The Setup
A wellness coach with years of live teaching and nothing to sell while she slept. Every workshop she ran was full of course-worthy material, and all of it evaporated the moment the session ended. She’d tried building an online course the manual way: rewatch recordings, transcribe, outline, write modules. Weeks of production for every hour of teaching. She quit halfway through twice.
The knowledge wasn’t the problem. The production line was.
The Build
I built her a pipeline instead of a process. Session recordings went in one end. Draft course modules came out the other.
Zapier watched for new recordings, sent them for transcription, and handed the transcripts to ChatGPT with prompts I wrote specifically for her teaching style — how she structures a concept, the exercises she assigns, the way she talks to students. The AI drafted module outlines, lesson summaries, and worksheets from what she’d already said out loud. She reviewed and polished instead of staring at a blank page.
The prompts were the real work. Generic summarization flattened her voice into wellness-blog mush. It took iterations to get output that sounded like her teaching, not like the internet’s average opinion about mindfulness.
The Mess
The first drafts were technically accurate and completely lifeless. That’s the trap with AI course production — it captures the information and drops the warmth, and for a coach the warmth is the product. We fixed it by feeding the system her actual language: recurring phrases, her framework names, the questions she asks students. The output stopped being a summary and started being a draft she recognized.
The other snag was trust. She’d been burned by tech promises before and expected to babysit every step. The pipeline had to prove itself a few times before she stopped re-checking its work.
The Result
Production time collapsed. The moment that sold it: she spontaneously put together a mini-course over a single weekend — something that would’ve taken weeks before. Not because she worked harder, but because the raw material was already drafted the moment she finished teaching it.
More courses shipped. And she got a piece of her energy back, because the part of the work she hated was gone and the part she loved — the live teaching — now did double duty.
The Takeaway
Experts don’t have a content problem. They have a capture problem. Automate the capture and the course already exists — it just needs editing.
Related: Course Creator Workflow Overhaul | Zapier + ChatGPT Business Automation