Building your first automation

Let's build something real. We'll walk through a classic first automation — AI email triage — because it uses every part of the model and delivers immediate value. Follow along in whatever tool you chose (the steps are the same everywhere).

The goal: when an email arrives, AI reads it, classifies it, and routes it — so your inbox sorts itself.

Step 1 — Trigger. Choose "New email" as the trigger and connect your email account (the platform handles the login). Now the automation fires whenever mail arrives.

Step 2 — AI step (classify). Add an AI step that takes the email's subject and body and classifies it. Your instruction to the AI might be: "Classify this email into one of: Sales lead, Support request, Billing, Personal, Spam. Also rate urgency: High, Medium, Low. Return just the category and urgency." This is the "classify" verb doing judgment a simple rule couldn't.

Step 3 — Logic (route). Add paths/branches based on the AI's output: if it's a "Sales lead," do one thing; if "Support," do another. This is deterministic logic reading the AI's decision.

Step 4 — Actions. For each path, take an action: a sales lead → create a task for the sales team and post to the #sales Slack channel; a support request → forward to the support inbox with the AI's summary. Map the fields by pointing and clicking.

Step 5 — TEST before you turn it on. This is the most important habit (Module 2 goes deeper): run the automation on a real sample email and inspect every step's output. Did the AI classify correctly? Did the right path fire? Did the action land? Read the actual output at each step — don't just trust it. Fix anything wrong before going live.

Step 6 — Turn it on. Once it works on test data, activate it. Now it runs automatically, sorting your inbox while you do other things.

Congratulations — you just built an automation that understands your email and acts on it, with no code. Notice how it used the whole model: a trigger, an AI step for judgment, logic to route, and actions to execute. Every automation you build is a variation on this.

Two beginner tips: start simple (get a basic version working before adding bells and whistles), and test with real data at every step (the number-one source of "why isn't it working?" is not checking each step's output). Build this, watch it run, and you've got the core skill — the rest of the course makes your automations more capable, reliable, and safe.