AI Classify

The AI Classify node uses AI to categorize input into predefined categories. It's perfect for sentiment analysis, intent detection, priority classification, and spam filtering.

Overview

AI Classify uses AI SDK v6 generateObject with enum output for reliable, structured classification. It guarantees the output will be one of your defined categories.

Configuration

Model Selection

Provider	Models	Credential Type
OpenAI	GPT-4.1, GPT-4.1 Mini, GPT-4o, o1, o3	OpenAI
Anthropic	Claude Sonnet 4, Claude 3.7, Claude 3.5 Haiku	Anthropic
Google	Gemini 2.5 Pro, Gemini 2.5 Flash, Gemini 2.0 Flash	Google AI

Settings

Field	Description	Default
Name	Node identifier for output reference	"AI Classify"
Model	AI model to use	gpt-4o-mini
Credential	Stored credential for the AI provider	Optional
System Prompt	Classification instructions	"Classify the input..."
Prompt	Direct text input	-
Input Variable	Variable from workflow context	-
Categories	Classification options	Positive/Negative/Neutral
Include Confidence	Add confidence score (0-1)	true
Include Reasoning	Add explanation for classification	false
Temperature	Creativity level (0-2)	0.1

Built-in Templates

Sentiment Analysis

Categories: positive, negative, neutral
Use Case: Customer feedback, reviews, social media

Intent Detection

Categories: question, request, complaint, feedback, other
Use Case: Customer support routing, chatbot classification

Priority Classification

Categories: urgent, high, medium, low
Use Case: Ticket triage, task prioritization

Spam Detection

Categories: spam, not_spam
Use Case: Email filtering, content moderation

Output

The node outputs structured classification data:

{
  "AI Classify": {
    "category": "positive",
    "label": "Positive",
    "confidence": 0.92,
    "reasoning": "The text expresses satisfaction with the product.",
    "usage": {
      "promptTokens": 80,
      "completionTokens": 25,
      "totalTokens": 105
    }
  }
}

Example Workflows

Customer Support Triage

[Webhook] → [AI Classify: Priority] → [Switch] → [Route to Team]

Configuration:

Categories:
  - value: urgent
    label: Urgent
    description: System down, security issue, data loss
  - value: high
    label: High
    description: Major feature broken, blocking issue
  - value: medium
    label: Medium
    description: Feature request, minor bug
  - value: low
    label: Low
    description: Question, documentation
Include Confidence: true
Include Reasoning: true

Sentiment-Based Response

[Form Input] → [AI Classify: Sentiment] → [Condition] → [Positive/Negative Response]

Content Moderation

[User Post] → [AI Classify: Spam] → [Condition] → [Approve/Reject]

Best Practices

Use low temperature - Classification benefits from deterministic outputs (0.1-0.3)
Write clear descriptions - Help the AI understand each category
Enable confidence scores - Useful for handling uncertain classifications
Use reasoning for debugging - Understand why classifications are made
Choose efficient models - GPT-4o-mini or Claude 3.5 Haiku work well for classification

Comparison with Other AI Nodes

Feature	AI Classify	AI Generate	AI Agent
Output Type	Structured enum	Free text	Free text
Tool Calling	No	No	Yes
Guaranteed Format	Yes	No	No
Best For	Categorization	Content creation	Complex tasks

AI Generate

Generate text content.

AI Agent

Complex AI tasks with tools.

Credentials

Manage API keys securely.

AI Generate

AI Agent

Credentials

On this page