Data Renegades: evergreen content pipeline
Built for Recce's Data Renegades podcast, an AI-assisted system that turns one recorded episode into a full slate of evergreen content.
A podcast episode is one hour of raw material. Repackaging it into everything an audience actually sees, LinkedIn posts, X threads, Reels, clips, carousels, a YouTube package, used to mean someone sitting down for the better part of a day. I designed the workflow and system that turns that into a 90-minute process.
What I did
I built the pipeline architecture: a transcript-analyst step that reads the raw episode and identifies the moments worth extracting, and a content-packager step that turns those moments into platform-ready assets. I designed the prompts, the structure, and the handoffs between steps, then wired it into the tooling (YouTube, Buffer, HubSpot) that gets the output published.
The numbers
4-8 hours down to 1.5 hours per episode, from raw recording to published content
~30 pieces of content per episode: 21 posts across LinkedIn, X, and Reels, 8 short clips, 1 carousel set, and 1 full YouTube package
Running consistently across 13+ episodes
Time saved that compounds
This wasn't a one-time script for a one-time favor, it's a system that pays out more the longer it runs. Every episode that goes through it banks the same 5-6+ hours back, and that saving compounds: 13 episodes in, it's already returned weeks of time that would otherwise have gone into manual repackaging, with no sign of slowing down as the show continues.
It's also worth naming when this was built. This wasn't a simple "summarize this transcript" prompt, the kind of remedial AI use that was common at the time. It was a structured, multi-step system with defined roles (a transcript-analyst step, a content-packager step) doing real judgment work, extracting the moments worth extracting and shaping them for specific platforms, built well before that kind of layered AI workflow was standard practice.
Why this one matters to me
This is the clearest proof I have that "infrastructure, not a deliverable" isn't just a pitch, it's how I actually build. The system runs without me reviewing every output, produces 30 pieces of content in the time it used to take to produce one, and the value keeps growing the longer it's in use. That's the whole idea behind GTM & Content Strategy work: build the thing that keeps paying off after you leave the room.