Course A · Module A3 No-Code Automation · Lesson 8 of 17

Build Your First Automation

Reading about automations is fine. Shipping one that runs without you is the thing that changes how you work.

~20 min · Have Zapier or Make open and logged in · Prereq: A7 (triggers & actions)

You understand triggers and actions. You've seen the template. Now you're going to build a real one — something that will run on its own after you close the laptop. This lesson's single win: by the end, you will have a live automation that watches for an event, does something useful with AI, and delivers a result — without you touching it. Small scope, real ship. That first run is the thing.

Pick your combination

Choose one row from each column. Simpler is better for your first build — you can layer complexity later. The goal is a working loop, not an impressive one.

Pick a trigger

New email in Gmail / Outlook
New form submission (Typeform, Google Forms)
New row in Google Sheets / Airtable

Pick an AI action

Summarize to 2–3 bullet points
Draft a reply in my tone
Classify as [category A / B / C]

Then deliver toPost to a Slack channel · Add a row in Google Sheets · Send an email

Build it — step by step

Start from a template, not a blank screen. In Zapier: go to Templates and search for your trigger app (e.g. "Gmail + ChatGPT" or "Typeform + Slack"). In Make: use the Template library. Pick the closest match and open it. You're modifying, not inventing.
Connect your trigger app. Authorize the app (Gmail, Sheets, etc.) and tell it which inbox, sheet, or form to watch. Test the trigger — Zapier/Make will pull in a real recent record so you can see the data flowing.
Set up the AI action. In the AI step, paste your prompt template from A5. Map the trigger fields into the brackets — e.g. replace [EMAIL BODY] with the trigger's "Body" field. Keep the prompt short and specific. One job per automation.
Set up the output action. Connect the destination (Slack, Sheets, email). Map the AI's output into the right fields — e.g. the AI's reply text goes into the Slack message body.
Test it with real data before turning it on. Use the built-in "Test" button to run the whole chain with a real sample. Read the output. Is the AI doing what you asked? Is it landing in the right place?
Turn it on. Watch the first live run. Enable the automation, then trigger it for real — send a test email, submit the form, add the row. Watch it run. Confirm the output landed. You're done.

Add a human-in-the-loop for your first version. Route the AI output to a Slack channel or a Google Sheet for review before it touches a customer. Once you've seen it run correctly a dozen times, remove the checkpoint if you're confident. Trust is earned in runs, not promised in setup.

Test before you trust

Before you declare it live:

Did the trigger fire on a real event? (Not just a test sample — an actual new email, form, or row.)
Did the AI output make sense? Read it. Would you have written something close to this yourself?
Did the output land in the right place? Right Slack channel, right sheet, right email address.
What happens if the input is weird? Try a short, blank, or unusual submission and watch what the AI does.

A zap that breaks on edge-case input isn't broken — it's unfinished. Fixing the prompt or adding a filter step is normal. Iteration is the job.

Check yourself

Answer before opening. Recalling it now is what makes it stick.

Why start from a template instead of a blank automation?

Templates wire up the trigger and action connections for you. The fiddly part of a first build isn't the AI prompt — it's connecting the apps correctly and mapping the right data fields. A template handles that, leaving you to focus on the prompt.

Your automation sends a drafted reply directly to the customer. It goes out wrong twice. What should you have built instead?

A human-in-the-loop checkpoint. Route the draft to a Slack message or a Sheets review queue first. Read it, approve, then send. Remove that checkpoint only after you've seen it run correctly enough times to trust it.

The AI action step produces a blank output on some runs. First thing to check?

Whether the trigger data field you mapped into the prompt is sometimes empty. If the prompt says "summarize [EMAIL BODY]" but some emails have no body (just a subject line), the AI has nothing to work with. Add a filter step that skips blank inputs, or update the prompt to handle them.