I run a podcast called Profitable Founder. I interview bootstrapped SaaS founders making real money. Twice a week. My only job is to sit down and have the conversation. Everything after recording is handled by AI agents.
- The Podcast Content Problem
- The Agent Team and Schedule
- Episode Detection (1 AM)
- Finding Clip-Worthy Moments (2 AM)
- Automated Video Editing (3:20 AM)
- Quality Gate (4 AM)
- Social Media Posting (5 AM)
- Newsletter Automation (2:30 AM)
- Analytics and Intelligence
- The Morning Brief
- Rules That Make It Work
- Results and Costs
The Podcast Content Repurposing Problem Every Creator Faces
Every podcast episode should generate: 3+ video clips for social media, posts for X with video attached, a newsletter teaser, YouTube SEO optimization, community content, performance analytics, and intelligence on trending topics.
Most podcasters do maybe 2 of these. Manually. Badly. Because there aren't enough hours.
So I built a team of AI agents to do all of them. Automatically. On a schedule. While I sleep.
The AI Agent Podcast Team: Who Does What and When
| Agent | Job | Schedule |
|---|---|---|
| Jimmy | Detects new episodes, downloads transcripts | 1 AM (Mon + Thu) |
| Claude | Picks 3 best clip moments, writes posts | 2 AM daily |
| Adrien | Edits video clips with subtitles | 3:20 AM daily |
| Bob | Quality checks everything | 4 AM daily |
| Dan | Schedules and posts to X, 3x/day | 5 AM daily |
| Tyler | Writes newsletter drafts | 2:30 AM (episode days) |
| Loop | Scrapes analytics from every platform | 6:30 AM daily |
| Crawly | Intelligence crawl: trends, competitors | 12:30 AM daily |
| Billy | Scans Skool for content gaps | 8 AM daily |
| Marc | Coordinates, delivers morning/evening briefs | 7 AM + 7:30 PM |
Automated Podcast Episode Detection with AI
Jimmy monitors the podcast RSS feed and YouTube Studio. When a new episode appears:
- Downloads the full transcript using the video-transcript-downloader skill
- Stores it in
transcripts/[episode-name].md - Logs metadata: title, guest, duration, YouTube URL, Spotify link
- Notifies Marc that new material is ready
One detection. Entire pipeline triggered. Every downstream agent picks up its work from this single event.
Using AI to Find the Best Podcast Clips Automatically
Claude reads the transcript and selects the 3 best moments:
- Concrete numbers: "We hit $2M ARR with zero funding"
- Contrarian takes: "I fired all my salespeople and revenue went up"
- Tactical insights: Specific how-to people can use immediately
- Emotional moments: Real vulnerability from founders
For each moment, Claude writes the X post (matching my writing style), provides exact timestamps, and verifies proper nouns. Everything goes to Notion. Claude tracks used clips in research/used-clips.md so we never repeat a moment.
Automated Podcast Video Editing with AI (Whisper + Smart Cuts)
Adrien picks up Claude's timestamps and produces broadcast-ready clips:
- Download: yt-dlp grabs the exact segment from YouTube
- Transcribe: Whisper generates word-level timestamps
- Smart cuts: Removes filler words automatically
- Silence removal: Tightens dead air
- Subtitle burn-in: White text, single line, 4-5 words max, synced to whisper
- Compress: h264, 16:9, 720p, under 16MB
Hard rules that took weeks to get right:
- Source clips MUST be h264 (VP9/AV1 breaks auto-editor)
- Never clip from the intro
- Clean start on complete thought, clean end on complete sentence
- No host voice bleeding at the tail
- Proper nouns verified: "Jenni AI" not "Jenny AI"
AI Quality Gate: Automated Content Review Before Publishing
Nothing goes live without Bob's approval:
- Subtitles match what the guest actually says?
- Timestamp correct (right moment, not 30 seconds off)?
- Writing style guide followed?
- Video clip attached? (Text-only: 11 views. Video: 1,500+.)
- Post starts with @mention? Reject. (X hides @mention-first posts.)
If something fails, Bob sends it back with specific notes. "Subtitle at 0:23 says 'Minea' but guest said 'Meenah'." Specific and actionable.
Automated X/Twitter Posting Schedule for Podcasters
Dan manages @profitfounder. Three posts per day for maximum timezone coverage:
| Time (Bali/SGT) | Target Audience |
|---|---|
| 4 PM | European evening scroll |
| 9 PM | US morning scroll |
| 4 AM | US afternoon break |
Workflow via Typefully API: upload video clip (3-step media upload), create draft with scheduled timestamp, post CTA reply after each post with YouTube + Spotify links.
Dan only posts from the last 4 episodes. When a new episode drops, the oldest falls off rotation. Keeps everything fresh.
AI Newsletter Automation: From Podcast Transcript to Beehiiv Draft
Tyler reads the transcript and drafts a newsletter. The absolute rule:
Newsletters are teasers only. Bullet points of what they'll learn. Never give away the content. Every bullet drives curiosity to click through to the episode. The biggest number or result goes in the subject line.
Tyler pushes to Notion. I review. I decide when it sends. Tyler drafts. I own the voice.
AI-Powered Podcast Analytics and Intelligence Gathering
Daily Platform Analytics (Loop, 6:30 AM)
Loop scrapes: YouTube views/subs/CTR, X followers/engagement per post, Instagram reach, Skool MRR/churn, Beehiiv open/click rates. Every Sunday: correlation analysis. Which clips drove follows? Which topics got engagement? What time slots perform best? This feeds back into Claude's moment selection.
Daily Intelligence Crawl (Crawly, 12:30 AM)
Crawly scans: 6 Reddit subreddits, Hacker News, competitor channels, X trends, Product Hunt launches. Structured intel brief hits Telegram by 12:45 AM. Strictly READ ONLY.
The AI Morning Brief: Everything You Need in One Telegram Message
Every morning at 7 AM, Marc sends me:
- What happened overnight (clips processed, posts scheduled, errors fixed)
- Intelligence highlights from Crawly
- Analytics snapshot from Loop
- Any issues Bob flagged
- Today's priorities
No checking dashboards. No logging into platforms. One message. Full picture. I wake up, read the brief, and know exactly where everything stands.
The Rules That Make Podcast Automation Work
Every AI Agent Needs an SOP (Standard Operating Procedure)
Each agent has sops/[agent-name].md: role, responsibilities, access, hard rules. Without SOPs, agents drift and overlap.
Strict Permission Matrix Prevents Disasters
Dan writes to X but can't delete Notion. Crawly and Billy are READ ONLY. Tyler writes drafts but can't send. Marc stays off social media entirely.
Master Rules Apply to All Agents
Every deliverable to Notion. No fake data. Write everything immediately. 80/20 only. Claude Opus everywhere.
Podcast Automation Results and Monthly Costs
| Metric | Result |
|---|---|
| YouTube subscribers | 7,600+ |
| X followers (@profitfounder) | 9,400+ |
| Skool MRR | $3,100/month (259 members) |
| Episodes archived | 26 with full transcripts |
| Content posted | 3x daily, every day |
| Cost Item | Amount |
|---|---|
| Mac Mini M4 | $700 one-time |
| Claude Opus API | $100-150/month |
| Typefully + Beehiiv | Free tier |
| Electricity | $3/month |
| Total monthly | Under $155 |
No video editor ($500-2000/mo). No social media manager ($1000-3000/mo). No newsletter writer. The entire operation runs for less than a single freelancer's day rate.
Lessons From 2 Months of AI Podcast Automation
- The sequence matters more than the agents. Jimmy → Claude → Adrien → Bob → Dan. Out of order = broken pipeline.
- Quality gates are non-negotiable. Skip QA and bad subtitles go live.
- Never automate the voice. AI writes drafts. I send the newsletter. I post in Skool.
- Track what works and feed it back. Loop's analysis tells Claude which moments performed. The pipeline improves itself.
- Clips without video are invisible. Text-only posts: 11 views. Video: 1,500+. No exceptions.
Get the Full Podcast Automation System
Inside OpenClaw Lab, members get this entire podcast pipeline as a template pack. Every SOP. The cron schedule. Writing style guide. Access matrix. Coordination playbook. Plus weekly lives where I walk through the system. 250+ founders building together.
New to OpenClaw? Start at installopenclawnow.com.
I share the exact playbooks, skill files, and workflows behind this system inside OpenClaw Lab. Weekly lives and AMAs with experts.
Join OpenClaw Lab →