AI Data Classification Agent in Make.com

This step-by-step guide shows you how to build an AI data classification agent in make.com that turns unstructured text into structured, actionable data using OpenAI and webhooks.

By the end, you will know how to configure a scenario, design an AI prompt, and parse structured responses so you can route messages, segment users, and trigger automated workflows.

What You Will Build with Make.com

In this tutorial, you will create an AI-powered data classifier that:

• Receives free-form text from a webhook trigger
• Uses an OpenAI model to classify and extract data
• Returns a structured JSON-like output
• Makes the data ready for routing, tagging, or storage

The agent runs entirely inside make.com, so you can easily connect it to CRMs, help desks, spreadsheets, or any other supported apps.

Prerequisites for Using Make.com

Before you start, make sure you have:

• An active make.com account with access to create scenarios
• An OpenAI API key or another compatible AI provider
• Basic understanding of JSON and key-value data structures
• Permission to send and process the text data you want to classify

If you need implementation or optimization help beyond this guide, you can work with automation specialists at Consultevo.

Overview of the AI Agent in Make.com

The data classification agent you build in make.com follows this general flow:

1. A webhook receives raw text or a payload from another system.
2. The scenario sends the text to an AI model with a carefully designed prompt.
3. The AI model responds with a strict JSON structure describing categories, entities, or tags.
4. Make.com parses the structured output and maps it to further modules.

To get started, study the reference walkthrough on the official how-to page for AI agents in make.com.

Step 1: Create a New Scenario in Make.com

The first step is to create a scenario that will host your AI agent logic inside make.com.

1. Log in to your make.com dashboard.
2. Click Create a new scenario.
3. Choose Webhook as the first module.
4. Select Custom webhook and create a new webhook URL.

Once the webhook is created, copy its URL. You will use this endpoint to send text for classification from external tools or test requests.

Configuring the Webhook in Make.com

In this stage, you want to make sure the incoming data can be mapped correctly.

1. Click the webhook module in make.com.
2. Enable Determine data structure if available.
3. Send a sample request to the webhook URL with the body that contains your text field (for example, message or content).
4. Allow make.com to detect the structure so you can reference this field later in the scenario.

Now your scenario can accept any text you want to classify.

Step 2: Add the AI Classification Module in Make.com

Next, you will add an AI module to perform the classification inside make.com.

1. Click the plus icon next to the webhook module.
2. Search for your AI provider, such as OpenAI.
3. Choose a Chat completion or Text completion module that supports system and user messages.
4. Connect your API key and select a suitable model (for example, a GPT-4 or GPT-3.5 family model).

This module will transform unstructured input into structured data.

Designing the Prompt in Make.com

The prompt is the heart of your agent. In make.com, configure the AI module with:

• System message that defines the agent role, such as: “You are a data classification assistant. Always return valid JSON.”
• User message that includes variables from the webhook, such as the incoming text field.

Tell the model exactly which keys and value formats you expect, and specify that the response must be valid JSON with no additional commentary.

Step 3: Define the Output Schema for Make.com

To make the output easy to use across modules, define a clear schema in your prompt. For example, you can ask the AI to return:

• category: A high-level label, such as “support”, “sales”, or “feedback”
• priority: A label like “low”, “medium”, or “high”
• entities: A list of key information (emails, order IDs, product names)
• summary: A short description of the message

In make.com, map the incoming text to the user prompt like this:

1. Use the webhook field (for example, {{1.content}}) in the user message.
2. Describe the required JSON structure explicitly.
3. Specify that missing fields should be null instead of omitted.

This gives you predictable outputs that are easy to parse inside make.com.

Example Prompt Structure in Make.com

Your user message might look logically like this (expressed in plain text to copy into make.com):

• “Classify the following message. Return valid JSON only with keys: category, priority, entities, summary. Message: <insert text from webhook field>”

Adapt this to the exact categories and fields your workflow requires.

Step 4: Parse and Use the Structured Output in Make.com

Once the AI returns structured data, you must parse it inside your scenario.

1. In the AI module response, identify the field that contains the JSON string.
2. Add a JSON > Parse JSON module in make.com.
3. Map the AI response text to the String input of the parse module.
4. Run the scenario once to let make.com detect the resulting JSON structure.

After parsing, each key from the AI output becomes a separate field you can map to other modules.

Routing by Classification in Make.com

To act on the classification results, add routers and conditional logic modules.

• Create a Router after the JSON parse module.
• Set conditions based on category or priority, such as:

• Route 1: category = "support"
• Route 2: category = "sales"
• Route 3: priority = "high"

Each route can send data to a different app or process within make.com, such as a help desk ticket, CRM record, or email notification.

Step 5: Test and Refine Your Make.com Agent

Testing is crucial to ensure your AI agent behaves consistently inside make.com.

1. Switch the scenario to On and send test requests to the webhook.
2. Observe the AI responses and parsed fields in the execution history.
3. Adjust your prompt if the JSON is invalid or inconsistent.
4. Add more examples to the prompt if the categories are unclear.

Iterate until the outcomes match your expectations across different sample texts.

Tips for Reliable Classification in Make.com

For more robust results, consider these guidelines:

• Use clear, concise instructions in the system message.
• Define all allowed categories in the prompt.
• Limit the response to a fixed schema to reduce hallucinations.
• Indicate how to handle unknown or ambiguous cases.

You can consult the official AI agent data classification guide on make.com for advanced prompt patterns and design recommendations.

Use Cases for AI Classification in Make.com

Once your agent is working, you can reuse the pattern across many workflows in make.com:

• Automatically triage support emails based on urgency and topic
• Route sales inquiries to the right team by region or segment
• Extract entities like order numbers or product IDs from messages
• Tag and store customer feedback in structured databases

Because all of this logic runs in make.com, you can plug the classifier into any existing automation scenario with minimal changes.

Next Steps

You now have a blueprint for building an AI data classification agent in make.com using webhooks, AI models, and JSON parsing. To deepen your skills, explore more advanced modules, combine multiple AI calls, and integrate with external systems for end-to-end automation.

For strategic guidance, architecture reviews, or complex multi-app workflows, you can partner with automation and AI experts via Consultevo while continuing to run your core logic in make.com.

Need Help With Make.com?

If you want expert help building, automating, or scaling your Make scenarios, work with ConsultEvo — certified workflow and automation specialists.

Get Help