Giving AI Hands

Automations turn AI from a "chat" into a "worker." This lesson gives you the full picture of how to connect AI to the tools you already use.

1. The Automation Platform Landscape

Understanding which tool to use when:

Zapier — The "iPhone" of automation

• Easiest to use, 6,000+ app integrations, visual workflow builder
• Best for: Non-technical users, simple 2-3 step workflows, high-volume simple automations
• Cost: the most expensive of the three at any given volume; a limited free tier exists
• Limit: Workflows over ~10 steps get expensive and hard to debug

Make.com — The "Pro" tool

• Materially cheaper than Zapier at the same volume, because it bills per operation rather than per task
• Infinitely more powerful: routers, iterators, data transformers, custom code
• Best for: Complex workflows (20+ steps), conditional logic, scaling
• Learning curve: Steeper. But once you learn it, you won't go back to Zapier.

n8n.io — The "Open Source" choice

• Self-hostable at no licence cost, or available as a paid cloud service
• Best for: Power users, data privacy concerns, custom AI agent orchestration
• Unique advantage: Can run Python/JS code inline, full API control
• Trade-off: Requires more technical knowledge

Which to choose?:

• New to automations → Start with Zapier for your first 3 workflows, then migrate to Make.com
• Want power + low cost → Make.com
• Privacy concerns or technical background → n8n

2. The Trigger-Action Loop

Every automation has two parts:

Trigger (The "When") — What starts the automation:

• New row in Google Sheet
• New email with specific subject
• New form submission
• New row in database
• Time-based (every hour, every day at 8 AM)
• Webhook received (from any app via HTTP)

Action (The "Do") — What happens when triggered:

• Send data to AI for processing
• Send email / Slack message
• Add row to spreadsheet
• Update CRM record
• Call a webhook
• Delay (wait 2 hours before next step)

Example — The Lead Research Bot:

1. Trigger: New row added to Google Sheet (columns: Company Name, Website, Contact Email)
2. Action: Perplexity searches for company's latest news and funding info
3. Action: OpenAI drafts personalized icebreaker email using the research
4. Action: Slack message to sales channel with: company info + research + draft email
5. Action: Human reviews and approves before sending

3. The Automation Map

Before you build, draw it. Two approaches:

• Box-and-arrow diagram: Draw boxes (steps) and arrows (flow direction). Label trigger, actions, decisions.
• The 3-question method:
  1. Where does data START? (The trigger — often a form, sheet, or email)
  2. Where does data END? (The destination — often an email, CRM, or dashboard)
  3. What TRANSFORMS it in the middle? (AI processing, formatting, filtering, routing)

4. Variables: The Secret to Scalable Automations

• Hardcoded (Bad): "Send email to [email protected]"
• Variable (Good): "Send email to [Email Address from Row 2]"
• Every piece of data that flows through your automation should be a variable, not hardcoded. This is what makes one automation serve thousands of leads.

5. Error Handling Basics

Every automation should have:

• Success path: The happy flow (everything works)
• Error path: What happens if AI fails, API is down, or data is missing
• Timeout: What if the AI takes too long?

Basic error handling: Wrap critical steps in a "If error → Alert me on Slack" branch.

⚠️ Common Mistakes

• Building too fast: Trying to automate a 20-step process on day one. Start with 2 steps.
• Ignoring the "Filter": Running an automation on every email instead of just the ones with the label "Sales". You will waste money on tokens!
• Not documenting: When it breaks (and it will), you'll want to know what you built. Take screenshots.
• No monitoring: Set up a daily check-in to see if your automations ran successfully. Silent failures are the worst kind.

What's Next: You know the tools. Let's build your first intake system using forms.