That thing you do every week by hand? Let’s make it happen while you’re not looking.
Overview
Install n8n, identify one repetitive task, build an automated workflow with AI processing, and test it with real data.
Draws from: Modules 07, 05 Time estimate: 2-3 hours Difficulty: Low-medium Prerequisites: A specific repetitive task you do weekly
Step-by-Step
Phase 1: Setup (15-30 min)
Cloud path: Sign up at n8n.io. You’re ready immediately.
Self-hosted path:
docker run -it --rm \
--name n8n \
-p 5678:5678 \
-v n8n_data:/home/node/.n8n \
docker.n8n.io/n8nio/n8n
Open http://localhost:5678.
Phase 2: Identify Your Workflow (15 min)
Pick one task from this list (or define your own):
| Task | Trigger | AI Processing | Output |
|---|---|---|---|
| Email digest | Schedule (daily) | Summarize + classify urgency | Telegram/Slack message |
| Content repurposing | New blog post (webhook) | Generate social variants | Buffer/scheduling tool |
| Invoice processing | New email with attachment | Extract key data | Spreadsheet row |
| News monitoring | RSS feed | Filter + summarize relevant items | Email or Slack digest |
| Customer inquiry routing | Form submission | Classify type + draft response | CRM + review queue |
Phase 3: Build (60-90 min)
Using the email digest example:
- Trigger node: Schedule Trigger → set to your preferred time
- Data source node: Email (IMAP) → configure with your email credentials → fetch last 24h
- AI processing node: Anthropic (Claude) or OpenAI → configure with API key → write the prompt:
Summarize these emails. For each: - Subject, Sender - Category: URGENT / NEEDS_RESPONSE / FYI - One sentence summary Group by category, URGENT first. - Output node: Telegram/Slack/Email → configure destination → format the message
- Test: Run manually. Check the output. Refine the prompt if needed.
Phase 4: Refine and Activate (30-45 min)
- Run the workflow 3-5 times with real data
- Adjust the AI prompt based on results:
- Is it classifying urgency correctly?
- Are summaries useful or too generic?
- Is the format readable?
- Add error handling (what happens if email fetch fails?)
- Activate the schedule
What Success Looks Like
- A working automated workflow that runs on schedule
- Real time savings on a task you were doing manually
- Understanding of the n8n node model well enough to build more workflows
Stretch Goals
- Add a second workflow for a different task
- Connect your local Ollama instance as the AI provider (saves API costs for high-volume workflows)
- Build a branching workflow with different paths based on AI classification
Previous: Project C: Set Up OpenClaw with One Real Skill | Back to Project Lab | Next: Project E: Build a Personal Tool