Building your first Make scenario step by step

Section 10.6.2 : Building your first Make scenario step by step

🎯 Learning objective

Build an end-to-end functional Make scenario integrating ChatGPT to automate a concrete business task: automatic form processing. You'll learn the complete process from trigger to logging, including conditional routing.

The project: automatic form processing

Imagine: you receive 50 requests per day through a contact form. Each request needs to be read, categorized (support? quote? complaint?), an appropriate response drafted, and everything logged in a tracking sheet. Manually, this takes 4 hours per day. With Make + ChatGPT, it takes 0 minutes — everything is automatic.

Here's the scenario we'll build:

1. →A Google Form is submitted → trigger
2. →Make retrieves the data → automatic extraction
3. →ChatGPT analyzes and categorizes the request → intelligence
4. →A Router sends it down the right path → conditional logic
5. →A personalized response email is sent → action
6. →Data is saved to Google Sheets → logging

Prerequisites: preparing your tools

Before building, prepare:

1. →A Google Form with 4 fields: Name, Email, Subject (dropdown: Technical Support / Quote Request / Complaint / Other), Message (long text)
2. →A Google Sheet "Request Tracker" with columns: Date | Name | Email | Subject | AI Category | Urgency | Summary | Status
3. →An OpenAI API key (from platform.openai.com → API keys)
4. →A Slack account (optional, for urgent notifications)

Step 1: Configure the trigger

Module: Google Forms > Watch Responses

1. →On the Make canvas, click the big + in the center → search for "Google Forms"
2. →Select the "Watch Responses" action
3. →Click "Add" to create a connection → authorize access to your Google account
4. →Select your form from the dropdown
5. →Limit: leave at 10 (max responses processed per execution)
6. →Click "OK" to confirm

What this module produces: on each execution, it returns new responses as structured data — {{1.Name}}, {{1.Email}}, {{1.Subject}}, {{1.Message}}.

Step 2: Analyze with ChatGPT

Module: OpenAI > Create a Chat Completion

Add the module by clicking the small + to the right of the trigger. Detailed configuration:

• →Connection: click "Add" → paste your OpenAI API key
• →Model: gpt-4o-mini (fast and affordable, ~$0.15/million tokens)
• →Messages:
  • →Role: System
  • →Message Content (the system prompt, crucial for quality):

You are a customer request processing assistant for a digital services company. You receive client messages and must analyze them accurately.

INSTRUCTIONS:
1. Categorize the request among: technical_support, quote, complaint, information, partnership
2. Assess urgency: high (blocking issue / unhappy customer), medium (standard request), low (simple question)
3. Summarize the request in ONE sentence
4. Write a professional, empathetic, personalized response (use the client's first name)

RULES for the response:
- Use informal tone if the client's message is casual, formal tone otherwise
- If complaint: start by acknowledging the problem and apologizing
- If quote: thank them and indicate a sales rep will follow up within 24h
- If support: provide a first troubleshooting suggestion if possible

RETURN STRICTLY as valid JSON, with no text before or after:
{
  "category": "...",
  "urgency": "high|medium|low",
  "summary": "...",
  "suggested_response": "...",
  "sentiment": "positive|neutral|negative"
}

• →Add a second message with Role: User and Message Content:

Name: {{1.Name}}
Email: {{1.Email}}
Subject: {{1.Subject}}
Message: {{1.Message}}

• →Temperature: 0 (for deterministic and consistent results)
• →Max Tokens: 500 (sufficient for the JSON response)
• →Response Format: select json_object (forces the model to return valid JSON)

Why gpt-4o-mini and not GPT-5? For this standard categorization/writing task, gpt-4o-mini is ~20x cheaper and sufficiently capable. GPT-5 is reserved for complex reasoning tasks. At 50 requests/day, the difference: $0.25/month (mini) vs $5/month (GPT-5). Over a year: $3 vs $60.

Step 3: Parse the AI's JSON response

The OpenAI module returns a JSON string in {{2.choices[0].message.content}}. To use individual fields (category, urgency, etc.) in subsequent modules, you have two options:

Option A — parseJSON inline (what we use):

{{parseJSON(2.choices[0].message.content).category}}
{{parseJSON(2.choices[0].message.content).urgency}}
{{parseJSON(2.choices[0].message.content).suggested_response}}

Option B — JSON > Parse module (cleaner for complex scenarios): Add a "JSON > Parse JSON" module between OpenAI and the Router. It creates properly named variables.

Step 4: Route based on category and urgency

Module: Router

Add a Router after the OpenAI module (or JSON Parse). Create 3 routes:

Route 1 — High urgency (complaint or blocking bug):

• →Condition: {{parseJSON(2.choices[0].message.content).urgency}} equals high
• →Actions:
  • →Slack > Send a Message → #client-alerts channel → message: 🚨 URGENT — {{1.Name}} ({{parseJSON(2.choices[0].message.content).category}}): {{parseJSON(2.choices[0].message.content).summary}}
  • →Gmail > Send an Email → immediate response (see Step 5)

Route 2 — Quote request:

• →Condition: {{parseJSON(2.choices[0].message.content).category}} equals quote
• →Actions:
  • →Gmail > Send an Email → thank-you email + "a sales rep will get back to you within 24h"
  • →Google Sheets > Add Row → add to CRM (sheet "Prospects")

Route 3 — Other requests (fallback):

• →Condition: none (default route)
• →Actions:
  • →Gmail > Send an Email → personalized standard response

Step 5: Send the response email

Module: Gmail > Send an Email

Configuration:

• →To: {{1.Email}}
• →Subject: Re: {{1.Subject}}
• →Content Type: HTML
• →Body:

<p>{{parseJSON(2.choices[0].message.content).suggested_response}}</p>
<br>
<p>Best regards,<br>The Support Team<br>
<em>This message was generated automatically. If you'd like to speak with a human, simply reply to this email.</em></p>

Important point: always add the "generated automatically" mention. It's a best practice both ethically and legally (AI transparency).

Step 6: Log to Google Sheets

Module: Google Sheets > Add a Row (present on ALL routes)

This Google Sheet becomes your dashboard: you can create charts to track volume by category, average sentiment, urgency trends, etc.

Step 7: Add error handling

This is the step beginners forget — and it's what separates a prototype from a production tool.

Error Handler on the OpenAI module:

1. →Right-click on the OpenAI module → "Add error handler"
2. →Choose "Resume" → the scenario continues despite the error
3. →Add a Gmail > Send an Email module in the handler: send yourself an alert email with error details
4. →Add a Google Sheets > Add a Row module: log the error in an "Errors" tab with the date, error message, and form data (for manual reprocessing)

Common OpenAI module errors:

• →429 Rate Limit Exceeded: too many simultaneous requests → add a Sleep module (2 seconds) before the OpenAI module
• →500 Internal Server Error: OpenAI outage → the error handler + automatic retry (configurable in scenario settings)
• →Invalid API Key: expired or revoked key → regenerate at platform.openai.com

Test the complete scenario

1. →Click Run once (green button at the bottom)
2. →Open your Google Form in another tab
3. →Submit 3 test forms with different profiles:
   • →Test 1: "My site has been down for 3 days, this is unacceptable" (high urgency, complaint)
   • →Test 2: "Hello, I'd like a quote for an e-commerce website" (quote, low urgency)
   • →Test 3: "Do you offer training courses?" (information, low urgency)
4. →Return to Make → verify all 3 bubbles are green (success)
5. →Click on each bubble to inspect input/output data
6. →Verify: email received? Slack notified (if high urgency)? Google Sheet updated?
7. →If everything works: activate scheduling → ON, interval: 15 minutes