Case 04·Content
Private conversations in. Publishable drafts out.
A content engine that turns every call, talk and recording into posts and newsletters in your voice, with privacy handled before a word is drafted. Proven on 800+ recordings.
Most businesses already hold their best thinking in recorded calls, talks and sessions, and almost none of it gets published. For this client, that was a library of 800+ private recordings, long and full of names. I built an engine that scores every recording for what's worth writing about, strips identifying details in four separate passes, drafts an SEO article and a newsletter version, runs an editor pass, and tells the owner when a batch is ready for review.
The problem
The raw material already existed. Turning it into content meant someone listening to hours of recordings, deciding what mattered, removing every private detail and then writing. That never happened.
How it works
Six components run end to end. The last step is a person.
- 01 Library Step
- 02 Score Agent
- 03 Anonymize Safeguard
- 04 Draft Agent
- 05 Editor pass Agent
- 06 Notify Step
- 07 Review and publish Human
Tap any stage to see what it does.
Press run to watch it move through every stage. When it reaches the human step, it waits for you.
The tradeoff
Privacy before speed. The four anonymization passes run before drafting, not after, so private details never reach the drafting model at all.
Where a person stays in charge
Nothing publishes on its own. The owner gets a batch notification and approves each piece.
By the numbers
recordings in the library
anonymization passes before any drafting
drafts from each selected recording: an article and a newsletter
Point it at any archive, from podcasts and webinars to client calls and talks.
Want this running on your business?
Live in 10 business days, shaped around your team.
Before and after
Sample runMaria Lopez said the turning point came when Harbor Lane Studio stopped discounting for Brightwave Retail. After moving to Austin in 2023, she raised prices by 18%...
Most founders treat a price increase as a sales problem. It usually starts earlier, with the clients they keep saying yes to.
Where else it fits
- Podcast and webinar archives turned into a year of content
- Sales-call libraries turned into objection-handling content
- Coaching and consulting sessions turned into newsletters, safely anonymized
- Internal talks turned into knowledge articles
- Conference talks and keynotes turned into articles and posts
- Course and training recordings turned into lead magnets
Questions
Does it need a specific AI model?
No. This build drafts with Gemini. I pick the model per job, and the same engine can run on Claude or OpenAI.
Our recordings contain sensitive client information. Is that a problem?
That's exactly why anonymization runs before drafting. Names and identifying details are removed in four passes, and nothing publishes without review.
Will it sound like us?
It drafts from your own recordings and real examples of your writing, and nothing publishes without your review.
Works with your tools
Shown here on
- Google Apps Script
- Google Drive
- Gemini API
Swaps in for what you already use: Microsoft 365, HubSpot, Salesforce, Pipedrive, Notion, Airtable, Slack or Teams, Zapier or Make.
Download the one-page summary (PDF)