How to build your first AI automation in Make.com (free plan, step by step)

Learn the 5 words Make uses (and what the free plan includes)

At the time of writing, Make's Free plan has no time limit and includes 1,000 credits a month, routers and filters, and a 15-minute minimum interval between scheduled runs. Free accounts can keep two scenarios switched on at once. Credits expire at the end of each month, and when they run out your scenarios stop until credits are added. Check Make's pricing page before you plan around a limit, because these details change.

Sketch the flow and prepare a test sheet

Every good automation starts on paper. Write it as one sentence:

When [a new form response arrives]
→ [AI sorts it into a category and writes a one-line summary]
→ [AI drafts a short reply]
→ [save everything to a "Triage" sheet]
→ I review and send replies myself.

Not sure what to automate first? Paste your idea into ChatGPT, Claude, Grok or Gemini and get the plan checked before you build:

I want to build my first automation in Make.com (free plan, 1,000 credits a month).
Idea: [describe it, e.g. "sort new feedback form responses and draft replies"].
1. Write it as: trigger → steps → output, using common Make modules (Google Sheets, Gmail, Make AI Toolkit, filters).
2. Estimate how many credits one run uses and how many runs per month I can afford.
3. List what could go wrong (empty fields, duplicates, AI mistakes) and how to guard against each.
4. Tell me which step must stay manual for safety.
Keep it short and practical.

Then set up the data. Create a Google Form with a few questions (for example Name, Email, Feedback). In the form's Responses tab, choose Link to Sheets so every answer lands in a spreadsheet. In that spreadsheet, add a second tab called Triage with the headers Date | Name | Email | Feedback | Category | Summary | Draft reply. Submit two or three fake responses so you have something to test with.

Create the scenario and add the trigger

1. Sign up for a free account at make.com, then click Create scenario (in some layouts it's Create a new scenario).
2. Click the big +, search for Google Sheets and choose Watch New Rows.
3. Click Create a connection and Sign in with Google. (Google Sheets connects directly. Gmail is different: personal @gmail.com accounts need a custom Google Cloud OAuth client to connect Gmail to Make, so we're leaving email out of your first build.)
4. Pick your spreadsheet and the form responses sheet, set Table contains headers to Yes, and set Limit to a small number like 5 while you test.
5. When Make asks Choose where to start, pick All (or Choose manually and select one test row), then save.

Click Run once in the editor toolbar. A bubble with a number appears above the module: that's how many rows it found. Click the bubble to see the data, with each column available by its header name. Seeing real output here makes the next step much easier.

Add the AI steps

Hover over the right side of the trigger, click the +, and search for Make AI Toolkit. Create a connection with Make's AI Provider (pick a small model; it uses fewer credits and is plenty for short text).

Module A: Categorize text. In Text to categorize, map the Feedback field from the trigger. Add categories with short descriptions, for example: Bug (something is broken), Feature request (asks for something new), Billing (payments, refunds, invoices), Praise (positive feedback), Other. Set Assign only one category to Yes.

Module B: Simple Text Prompt. Add another Make AI Toolkit module and write the prompt in the Text field, mapping the form fields where the brackets are:

You help a small business triage customer feedback.
Customer name: [map Name]
Feedback: [map Feedback]

Return exactly two lines and nothing else:
SUMMARY: one sentence, under 20 words, saying what the customer wants.
REPLY: a friendly 2-3 sentence reply that thanks them by first name and says what happens next. Don't promise refunds, fixes, dates or discounts. If the feedback is unclear, ask one short question instead.

Click Run once again and open the bubbles on the AI modules. Read the output critically: is the category right, is the reply something you'd actually send? Adjust the category descriptions or the prompt and re-run. You'll usually need two or three rounds; that's normal.

Save the results and add a filter

Add a final module: Google Sheets → Add a Row. Choose the same spreadsheet and the Triage tab, then map each column: Date, Name, Email and Feedback from the trigger, Category from the Categorize text module, and the AI text from Simple Text Prompt for Summary and Draft reply. (If you want them in separate cells, you can split the two lines later with a formula in Sheets, or run two smaller prompts.)

Next, protect your credits with a filter: click the line between the trigger and the first AI module and set up a filter so the scenario only continues when Feedback exists (isn't empty). Test responses with an empty answer then stop at the filter instead of spending AI credits.

Run once more, then open the Triage tab. You should see a new row for each test response with the category, summary and draft reply filled in. If something is missing, click the bubble on the failing module: the error message and the input it received usually show exactly which field wasn't mapped.

Schedule it sensibly, turn it on, and keep a human in the loop

Click the schedule setting in the editor toolbar (it shows Every 15 minutes by default). This choice matters more than anything else on the free plan, because a polling trigger like Watch New Rows uses a credit every time it checks, even when nothing is new. Do the math before you switch it on:

For a feedback inbox, twice a day is usually plenty. Choose Daily (you can add several run times) or Weekdays, save, and switch the scenario on. Check your credit usage in Make after the first few days.

Then set a few house rules:

• Drafts, not sends. This scenario writes replies into a sheet; you read and send them. Only automate sending once you've reviewed dozens of drafts and trust the pattern, and never for refunds, legal, medical or angry-customer messages.
• Mind the data. Form answers go through Make and its AI provider. Make says AI Toolkit data isn't used to train the AI, but don't route passwords, payment details or health information through any automation you haven't cleared with whoever owns that data.
• Watch for failures. Make shows each run in the scenario's history. If a connection expires or a step errors, fix it there; a scenario that quietly stopped is the most common automation problem.
• Grow one step at a time. Once this works, try variations: a router that sends "Bug" rows to a different tab, a notification to Slack or email for "Billing", or a weekly summary of all feedback. Make also has a template gallery to borrow from. If you hit the two-active-scenario limit or need more credits, that's the point to compare paid plans.

Building your first automation is mostly about the thinking, not the tool. For ideas on what to automate next, see our guide to automating repetitive busywork with AI, and for formula help with your new Triage sheet, AI for Excel and Google Sheets.