← All publications
AI Scans 400 Meetings and Writes Posts 5/6
400 meeting summaries in Obsidian. Real client numbers. Concrete results. Proven methodologies.
The problem: impossible to re-read everything to extract content.
The solution: Claude Code with parallel agents.
🤖 The exact process
I launched 3 agents in parallel in Claude Code, each with a specific mission:
• Agent 1 → Scan meetings for SEO insights
• Agent 2 → Scan for AI/automation insights
• Agent 3 → Scan for business/entrepreneurship insights
Each agent read 20-35 meeting summaries in depth. They were looking for: precise numbers, client case studies with results, original methods.
Total time: ~5 minutes to scan 400 files.
📊 What the agents found
Hidden gems buried in my meetings:
• "120,000 → 240,000 impressions in 6 weeks" (e-commerce client)
• "30x impressions on a 79-page site" (cosmetics client)
• "13,000 → 50,000 clicks/day on Google Discover" (media client)
• "3,500 product pages rewritten in 2 hours for $500" (Shopify client)
These numbers come from real conversations. That's what makes the posts credible.
✍️ From insight to LinkedIn post
For each insight, I asked Claude Code to generate a structured post:
1. Hook (punchy first line with a number)
2. Context (the problem)
3. Solution (with emojis + sections)
4. Measurable result
5. Engagement question
I reviewed, adjusted the tone, validated. No publishing without human review.
📝 The workflow in Obsidian
All posts live in a single Markdown file. When I like a post, I add "OK" after the title. Claude Code schedules it automatically and marks it "SCHEDULED DD/MM".
15 posts written in one day. All based on real data.
Last post in the series: automating LinkedIn publishing in 2 languages.
How much untapped content is sitting in your meeting notes?
The problem: impossible to re-read everything to extract content.
The solution: Claude Code with parallel agents.
🤖 The exact process
I launched 3 agents in parallel in Claude Code, each with a specific mission:
• Agent 1 → Scan meetings for SEO insights
• Agent 2 → Scan for AI/automation insights
• Agent 3 → Scan for business/entrepreneurship insights
Each agent read 20-35 meeting summaries in depth. They were looking for: precise numbers, client case studies with results, original methods.
Total time: ~5 minutes to scan 400 files.
📊 What the agents found
Hidden gems buried in my meetings:
• "120,000 → 240,000 impressions in 6 weeks" (e-commerce client)
• "30x impressions on a 79-page site" (cosmetics client)
• "13,000 → 50,000 clicks/day on Google Discover" (media client)
• "3,500 product pages rewritten in 2 hours for $500" (Shopify client)
These numbers come from real conversations. That's what makes the posts credible.
✍️ From insight to LinkedIn post
For each insight, I asked Claude Code to generate a structured post:
1. Hook (punchy first line with a number)
2. Context (the problem)
3. Solution (with emojis + sections)
4. Measurable result
5. Engagement question
I reviewed, adjusted the tone, validated. No publishing without human review.
📝 The workflow in Obsidian
All posts live in a single Markdown file. When I like a post, I add "OK" after the title. Claude Code schedules it automatically and marks it "SCHEDULED DD/MM".
15 posts written in one day. All based on real data.
Last post in the series: automating LinkedIn publishing in 2 languages.
How much untapped content is sitting in your meeting notes?