How to create AI agents with Make: flows that make decisions, extract data, send emails and act on their own without writing code

Sign up for Make:

1. Go to https://www.make.com and click “Get started free”.
2. Create an account with your email or Google.
3. Choose a plan: the free plan works for exploring; for active AI agents you need Core ($9/mo). Confirm current options on the official page.

What is an AI agent vs. a regular scenario:

• A traditional scenario follows fixed rules: “when an email arrives with subject X, save the attachment to Drive”. It’s predictable but rigid.
• An AI agent has a goal in natural language (“process new orders”) and decides on its own how to achieve it: calls tools, analyzes responses, makes decisions along the way.

Maia, Make’s internal agent, can make decisions without needing an external API. For more control, you’ll connect Claude or ChatGPT in Step 3.

If you already know Make but have never used agents, check the tutorial How to automate tasks with Make (Integromat) and AI to refresh the basic concepts of scenarios and modules.

Before touching Make, take 5 minutes to answer these three questions. This way the agent will be clear and useful from the start.

1. What triggers it? Examples:

• A new email arrives in my inbox
• A Typeform form is submitted
• A new order comes in on Shopify
• A fixed schedule (every Monday at 9am)

2. What does the AI need to decide? Examples:

• “Is this email a complaint or a sales inquiry?”
• “Is the order complete or is information missing?”
• “What’s the best response for this customer?”

3. What action does it execute after deciding? Examples:

• Send a reply email
• Update a row in Google Sheets
• Create a task in Notion or Trello
• Send a Slack notification

Concrete example for this guide:

Trigger: new email with the word “order” AI decision: extract customer name, product and address from the email body Action: add a row to Google Sheets with that data and send a confirmation email to the customer

With this clear map you can build the scenario in Make in under 30 minutes.

Create the scenario:

1. In Make, click “Scenarios” in the left menu and then the ”+” button.
2. Click the center circle and search for your trigger module (for example, Gmail > Watch emails).
3. Configure the trigger: choose the Gmail account, the filter (subject contains “order”) and the check frequency.

Add the AI module (Claude or ChatGPT):

1. Click the ”+” that appears after the trigger module.
2. Search for “Anthropic” for Claude or “OpenAI” for ChatGPT.
3. Select “Create a completion” (or the equivalent module).
4. Connect your account: Make will ask for your provider’s API key.
   • Get your Claude API key at: https://console.anthropic.com
   • Get your OpenAI API key at: https://platform.openai.com/api-keys
5. In the “Prompt” field, write the instruction for the AI. You can use variables from the email:

Extract from the following email:
- Customer name
- Product requested
- Delivery address

Email: {{1.body.text}}

Reply in JSON format: {"name": "...", "product": "...", "address": "..."}

6. Click “OK” to save the module.

Using JSON in the response makes it easy for the following modules to read the extracted data without ambiguity.

If you want to understand what an AI agent is from scratch, check the tutorial How to create your first AI agent without coding (2026).

With the AI connected and extracting data, add the action modules that do the real work.

Action 1: Add a row to Google Sheets

1. Click the ”+” after the AI module.
2. Search for “Google Sheets > Add a row”.
3. Connect your Google account and choose the spreadsheet.
4. Map the fields: in the “Name” column place the variable {{2.name}} (the JSON response from the AI), in “Product” place {{2.product}}, etc.

Action 2: Send a confirmation email

1. Add another module: “Gmail > Send an email”.
2. In the “To” field, use the original sender’s email: {{1.from.email}}.
3. Write the subject and body. You can also ask the AI to draft the reply by adding another Anthropic/OpenAI module call with the prompt:

Write an order confirmation email in English, warm and professional, for customer {{2.name}} who requested {{2.product}}. Maximum 4 lines.

Adding decision logic (Router): If you want the agent to take different paths (complaint email vs. sales inquiry), add a “Router” module between the AI module and the actions. Each branch of the router corresponds to a different decision.

Before activating the scenario to run on its own, test it with real data.

Manual test:

1. In the scenario editor, click the “Run once” button.
2. Check that the trigger fires (Make will look for real emails matching your filter).
3. Watch the flow on the canvas: each module shows a bubble with the number of records processed. Click the bubble to see the exact data that passed through each module.
4. If something goes wrong, click the module with the error to see the message and fix it.

Activate the scenario:

1. When the test is successful, flip the switch in the bottom-left corner from “Off” to “On”.
2. Make will run the scenario at the frequency you set (every 15 minutes, every hour, at 9am, etc.).

Monitor the history: Go to “History” in the scenario menu to see each run: how many records it processed, if there were errors and how many operations it consumed.

Problem	Fix
AI module returns a 401 error	Your API key is incorrect or expired. Generate a new one at console.anthropic.com or platform.openai.com
AI returns text instead of JSON	Adjust the prompt: add “Reply ONLY with the JSON, no additional text” at the end
Google Sheets says “Permission denied”	Make sure you connected the correct Google account and the sheet is shared with that account
Trigger doesn’t detect new emails	Check the subject filter, change the check frequency to “immediately” or test with a test email
Free plan operations run out	Consider upgrading to Core ($9/mo) or reduce the scenario frequency to every hour instead of every 15 minutes