The Daily Grind: How We Automated Our Own Content Pipeline with n8n and Claude
We build automated workflows for clients. We also use them ourselves. The Daily Grind is our internal content automation system: five n8n workflows, one orchestrator, and Claude doing the heavy lifting.
The Problem
Running an AI agency means constantly creating content, prospecting, and staying sharp technically. Doing all three consistently is hard when you're also shipping client work. The bottleneck isn't ideas. It's the daily friction of sitting down and doing it.
We decided to automate what we could.
Five Workflows, One Trigger
The orchestrator fires at 5pm every day via cron. It kicks off four sub-workflows in sequence:
- Content Engine - Claude generates a 30-60 second video script, a LinkedIn post, and an X post around a single topic.
- Prospect Finder - Claude surfaces a batch of businesses that would benefit from AI automation, then logs them directly to our Notion pipeline.
- Shooting Content - Claude generates video shoot ideas and b-roll concepts based on what we've been building.
- Skill Move - Claude recommends one specific skill, tool, or workflow pattern to learn or practice that day.
Each sub-workflow calls the Claude API directly via HTTP node, parses the JSON response, and hands structured data downstream.
The Part That Makes It Smart
The orchestrator reads from n8n's workflow static data before firing. It passes recent topics, recent shooting ideas, and recent skills to each sub-workflow. Claude uses that context to avoid repetition across days.
The content engine also alternates style based on day of week. Odd days are Builder mode: show what we built. Even days are Teacher mode: explain how AI solves a real problem. This keeps the feed from going stale.
What Claude Gets as Context
The content engine system prompt is opinionated. It references real projects we've shipped: a custom MCP server pulling live FRED fertilizer price data for an ag client, an AI soil asset valuation portal built in Next.js and Supabase, a tactical operations SaaS for SWAT teams, an AI report-writing tool for law enforcement.
That specificity matters. Generic "AI is changing business" content doesn't land. Content grounded in real builds does.
What We Got Out of It
Each evening we get: a ready-to-record video script, a LinkedIn post, an X post, a refreshed prospect list in Notion, and one concrete skill to work on. None of it requires us to sit down and think from scratch.
The prospect finder in particular has changed how we approach outreach. Instead of searching manually, we're working a list Claude surfaces daily, already filtered by industry fit and AI readiness.
The Build
The whole suite is about 500 lines of JSON across five workflow files. Total build time was a few hours. The orchestrator pattern, one parent workflow calling modular children with shared context, is something we use in client builds too. It keeps each piece testable and replaceable.
If you're doing any volume of outreach, content, or skill-building, the automation exists to do the repetitive parts. We eat our own cooking here.