The short answer to “how to use Make with AI automation” is this: pick a repetitive task, connect Make to an AI model like OpenAI or Claude, and let the scenario run on a trigger. Make (formerly Integromat) is a visual automation platform that lets you chain apps, APIs, and AI together without writing code. You build scenarios by dragging modules onto a canvas, mapping data between them, and setting triggers like a new email, a form submission, or a scheduled time. The AI piece comes in when you add a module that calls a large language model, processes its output, and routes the result to another tool. That’s the whole loop: trigger, AI action, output delivery.

Why AI Automation With Make Matters for Business

Most businesses are sitting on a pile of small tasks that eat hours every week. Someone is copying data from a CRM into a spreadsheet. Someone else is summarizing customer emails. A third person is drafting the same type of response over and over. These tasks are perfect for AI automation because they follow a pattern but still need a brain to handle the variation.

Make is the right tool for this because it sits between your apps and your AI models. You can pull a row from Airtable, send it to GPT-4 for a written summary, then push the result into Slack or Notion. All of this happens on a visual canvas where you can see every step. No code, no server, no DevOps team.

The business case is straightforward. A workflow that takes a person 20 minutes and runs 10 times a week saves you over three hours a week. Multiply that across a team of five and you’re looking at more than 15 hours back per person every month. That’s time your team can spend on work that actually requires judgment.

What makes Make different from older tools like Zapier is the visual builder and the pricing model. You can see the entire scenario as a flowchart, which makes debugging much easier. Operations teams can hand a Make scenario to a colleague and that person can follow the logic without learning a programming language.

Step-by-Step: How to Use Make With AI Automation

The best way to learn is to build something real. I’ll walk you through a concrete example: a scenario that watches a shared inbox, uses AI to classify each email by intent, and routes it to the right Slack channel. You can adapt this pattern to almost any workflow.

Step 1: Create a Make Account and Open a New Scenario

Go to make.com and sign up. The free tier gives you 1,000 operations per month, which is enough to test a few scenarios. Once you’re in, click “Create a new scenario” in the top right. You’ll see an empty canvas with a question mark in the middle. That question mark is your first module.

Step 2: Set Up the Trigger Module

Click the question mark and search for the app you want to watch. For this example, pick Gmail. Choose the trigger “Watch emails” and connect your Google account. Make will ask for permission to read your inbox. Once connected, set the folder to watch (usually “INBOX”) and choose a filter if you want to limit which emails trigger the scenario.

A trigger is what kicks off your scenario. Without it, nothing runs. Common triggers include new rows in a database, new form submissions, scheduled times, and webhooks from other apps.

Step 3: Add the AI Module

Click the plus icon on the right edge of your Gmail module and search for “OpenAI” or “Anthropic”. Make has native modules for both. Pick the one that matches your account. For OpenAI, choose “Create a chat completion” or “Create a completion”. For Anthropic, choose “Create a message”.

In the module settings, you’ll need to map data from the Gmail trigger into the prompt. Click on the message field and drag in the email subject and body from the Gmail module. Make will show you the variables as blue bubbles. Your prompt might look like this:

“Classify the following email into one of these categories: sales, support, billing, or other. Reply with only the category name. Email subject: {{1.subject}} Email body: {{1.body}}”

Set the model to GPT-4o or Claude 3.5 Sonnet depending on which provider you picked. Keep the temperature low, around 0.2 to 0.3, for classification tasks where you want consistent answers.

Step 4: Add a Router to Split the Flow

Now you need to send each email to the right Slack channel based on the AI’s classification. Add a Router module from the flow control tools. A router lets you create multiple branches that run in parallel based on conditions.

Create four routes: one for sales, one for support, one for billing, and one fallback. In each route’s filter, set the condition to check the AI module’s output. For the sales route, the filter would be “AI response text contains sales”. Repeat for the other categories.

Step 5: Connect Slack to Each Route

Inside each route, add a Slack module. Choose “Create a message” and connect your Slack workspace. Map the AI’s classification into the message text along with the original email subject and sender. Pick the right channel for each route: #sales-inbox for sales, #support-inbox for support, and so on.

Step 6: Test the Scenario

Before you turn it on, run the scenario once using the “Run once” button at the bottom of the canvas. Make will fetch a real email from your inbox, send it to the AI model, and post the result to Slack. Watch each module light up green as it completes. If a module turns red, click on it to see the error message. Most errors come from bad mappings or expired API credentials.

Step 7: Schedule the Scenario

Once testing works, click the scheduling toggle at the bottom. Set it to run every 15 minutes for near real-time processing, or every hour if you don’t need instant routing. Make will execute the scenario on its servers, not yours, so your laptop can be off.

Step 8: Add Error Handling

Click on any module and look at the error handling settings. Set up a fallback route that sends a notification to a private Slack channel if something fails. You can also add an “Iterator” module if you need to process batches of items, or a “Data Store” module if you want to keep a record of what the AI classified.

That’s the full pattern. Trigger, AI action, router, output. Once you’ve built this once, you can apply the same shape to dozens of other workflows.

Common Mistakes and How to Avoid Them

The first mistake people make is using too high a temperature for factual tasks. If you’re asking the AI to classify, extract, or summarize, set the temperature to 0.2 or 0.3. Higher temperatures introduce randomness, which is fine for creative writing but bad for structured data work.

The second mistake is stuffing too much into a single prompt. If you need the AI to do five things, break it into five modules or five separate calls. Each call should have one clear job. This makes debugging easier and lets you swap models without rewriting the whole scenario.

The third mistake is forgetting about cost. Every AI call costs money, and Make charges per operation. A scenario that runs every minute and makes three AI calls per email can burn through your budget fast. Add a filter at the trigger level to only process emails that match certain criteria, and batch items where possible.

The fourth mistake is not logging AI outputs. Always store the AI’s response in a data store or a Google Sheet so you can audit it later. LLMs hallucinate, and you need a way to check what the model actually said versus what you expected. This is also useful for fine-tuning prompts over time.

The fifth mistake is treating Make as a replacement for a database. Make has data stores, but they’re not designed for large-scale relational data. If your workflow needs to join tables or run complex queries, use Airtable, Notion, or a real database and let Make move data in and out.

The sixth mistake is ignoring rate limits. Both Make and AI providers have rate limits. If your scenario fires too often and hits a limit, it will fail. Add a sleep module between batches or use Make’s built-in throttle settings to stay under the cap.

The seventh mistake is building a scenario that’s too complex on day one. Start with the simplest version that works, then add branches, filters, and error handling one at a time. A working simple scenario beats a broken complex one every time.

Where to Go From Here

Once you’ve built your first AI workflow in Make, the next step is to map out the other repetitive tasks in your business. Look for anything that involves reading text, making a decision, and writing a response. Those are the prime candidates for AI automation.

A useful framework is to ask three questions for every task. First, does this task follow a pattern most of the time. Second, can a language model make the decision a human is currently making. Third, what tool receives the output. If you can answer all three, you have a workflow worth automating.

For a deeper look at how to structure AI across your operations, the guide below covers the full operating layer, from prompt design to governance.

Free download: The AI Operating Layer We put together a practical guide covering this and more. Download it here.

For a structured walkthrough of building this into your operations, book a 60-min Omni Audit — https://calendly.com/sam-mckay/discovery-call?utm_source=edna-landing&utm_medium=blog&utm_campaign=product-keywords