Automate Your Inbox with ChatGPT and Make: A Step-by-Step Tutorial

Most inboxes don't need a smarter filter. They need a summary. This tutorial builds a no-code automation that watches a mailbox, sends new emails to ChatGPT for a plain-language summary, and delivers the result wherever you want it, a Slack channel, a daily digest email, or a spreadsheet row. The tool doing the connecting is Make (formerly Integromat), a visual automation platform that talks to hundreds of apps without requiring a developer.

What you'll need

• A Make account (the free tier covers this project)
• An OpenAI API key with a small amount of credit loaded
• Access to the mailbox you want to automate, Gmail or Outlook both work with Make's native connectors

Step 1: Set up the trigger

In Make, create a new scenario and add your email provider's "Watch emails" module as the trigger. Configure it to watch a specific folder, your main inbox works, but a filtered label like "To Summarize" keeps the automation from firing on every promotional email you receive. Set the polling interval to however often you want new mail checked, every 15 minutes is a reasonable default.

Step 2: Connect the OpenAI module

Add an OpenAI (ChatGPT) module immediately after the trigger. Select "Create a Chat Completion" as the action. This is where the actual summarization happens, and the quality of the result depends entirely on the prompt you configure here.

A prompt that works well for this use case:

"Summarize the following email in 2-3 sentences. Identify whether it requires a response, and if so, by when. Output format: Summary: [text]. Action needed: [yes/no, deadline if any]."

Map the email's subject and body fields from the trigger module into this prompt as variables, so each incoming email gets processed individually.

Step 3: Route the output

What happens next depends on your workflow. Three common options:

• Slack digest: Add a Slack module that posts the summary to a private channel, giving you a scannable feed instead of an inbox you have to open.
• Daily email digest: Instead of sending each summary immediately, use Make's aggregator function to collect them and send one consolidated email at a fixed time each day.
• Spreadsheet log: Add a Google Sheets module that appends each summary as a new row, useful if you want a searchable record over time.

Step 4: Test before you trust it

Run the scenario manually against a handful of real emails before switching on the scheduler. Check specifically for two failure modes: the summary missing an action item that was actually present, and the automation firing on emails you didn't want processed at all, like automated notifications. Both are fixable by refining the prompt or tightening the trigger's folder filter.

Securing the automation

Email content is sensitive by default. Keep the OpenAI API key restricted to this single use with usage limits set in the OpenAI dashboard, and avoid routing anything containing financial or health information through a third-party automation without checking your organization's data policy first. Make encrypts credentials at rest, but the safest approach is still to scope what the automation can see in the first place, via the folder filter from Step 1, rather than relying on encryption alone.

Where to take it from here

Once the basic pipeline works, the same pattern extends easily: route urgent emails to a different Slack channel than routine ones, have ChatGPT draft a suggested reply alongside the summary for your review, or trigger a calendar event automatically when an email mentions a meeting request. The hard part, wiring the trigger to the AI step, is already done. Everything after that is composition.