A beginner's guide to AI automation with Zapier and Make

What "automation" actually means

Every automation follows the same shape: trigger → action(s). Something happens (a new email arrives, a form is submitted, a spreadsheet row is added) — and in response, a series of steps runs automatically. Add an AI step in the middle and you get workflows that read, summarise, categorise or reply on your behalf.

Zapier vs Make: which one to start with

Both do the same thing. The difference is style:

• Zapier is simpler. Linear step-by-step builder, more app integrations, and the free tier is enough to learn on. Start here if you've never automated anything.
• Make is more visual and more powerful. You can branch, loop, and handle errors in ways Zapier can't. Move here once you've hit Zapier's limits — usually after a month or two.

Your first automation: AI-powered inbox triage

A workflow anyone can benefit from — sort incoming email by importance with AI, and get a Slack ping only for the urgent ones.

Step 1: The trigger

In Zapier, create a new Zap. For the trigger, pick Gmail → New Email. Connect your account and choose the inbox label to watch (Inbox is fine to start).

Step 2: The AI step

Add a step: ChatGPT / OpenAI → Send Prompt. Paste this prompt (Zapier will let you insert the email subject and body as variables):

Classify the email below as one of: URGENT, NORMAL, NEWSLETTER, SPAM. Reply with only the single word.
---
Subject: {{email_subject}}
Body: {{email_body}}

Step 3: The filter

Add a Filter step: only continue if the ChatGPT output equals "URGENT". This is what prevents Slack from lighting up with every newsletter.

Step 4: The action

Final step: Slack → Send Direct Message. Post to yourself with the subject, sender and a link to the email. Test it, turn the Zap on, and your inbox now has a triage layer running 24/7.

Three more automations worth building

1. Form submissions → cleaned CRM entry

Trigger: new Typeform / Google Form response. AI step: extract name, email, company and intent into structured fields. Action: create a row in Airtable or a contact in HubSpot. Saves the manual copy-paste every sales team wastes hours on.

2. Meeting notes → project tracker

Trigger: new file in a Google Drive "Transcripts" folder. AI step: extract action items in the format "Owner — Task — Due". Action: create tasks in Notion or Asana. Turns your weekly meeting into a populated task board with zero manual entry.

3. Support ticket → smart reply draft

Trigger: new email tagged "support". AI step: given your product knowledge base as context, draft a reply. Action: save the draft in Gmail (not send — always keep a human in the loop for customer replies). Cuts response time from hours to minutes.

The rules that keep automations from breaking

• Always start with a filter. A trigger that fires too broadly will burn through your free-tier task quota in a day.
• Keep a human step for anything that reaches a customer. Draft, don't send. Suggest, don't decide.
• Name your automations clearly. "Gmail → AI → Slack" means nothing in six months. "Urgent email triage v2" does.
• Log to a spreadsheet. Add a Google Sheets step at the end that appends what happened. When something misbehaves, you have a record to debug from.

What to build next

Once one automation is running, look for anywhere you do the same thing more than three times a week. That's almost always automatable. The compounding effect is real: five automations running in the background can quietly claw back a full day of your week — every week — for as long as they keep running.