AI-Powered Workflows

Use AI models to add intelligence to your automations — from content generation to data analysis and decision-making.

Use Case Examples

  • Generate product descriptions from raw specifications
  • Summarize long documents or emails
  • Extract structured data from unstructured text
  • Make decisions based on natural language input

Step 1: Add an AI Node

  1. Open the workflow builder
  2. Add an AI Chat Model node from the node palette
  3. Select your preferred AI model

Step 2: Write Your Prompt

The prompt is the instruction you give to the AI model. Write it clearly and include the data you want processed.

Example: Content Generation

Write a product description for the following item:

Name: {{trigger.product_name}}
Features: {{trigger.features}}
Price: {{trigger.price}}

Keep it under 100 words. Use a professional tone.

Example: Data Extraction

Extract the following fields from the text below and return as JSON:
- name
- email
- phone
- company

Text: {{trigger.body}}

Example: Summarization

Summarize the following email in 2-3 bullet points:

{{gmail.body}}

Step 3: Use the AI Output

The AI node outputs the model’s response as text. You can:

  • Map it directly into another node (e.g. Gmail body, CRM field)
  • Parse it with a JSON node if you asked the AI to return structured data
  • Branch on it with a Flow Control node for decision-making

Step 4: Chain Multiple AI Steps

You can use multiple AI Chat Model nodes in sequence for complex tasks:

  1. AI Node 1 — Extract key information from raw input
  2. AI Node 2 — Generate a response based on the extracted data
  3. AI Node 3 — Translate the response to another language

Each node can use the output of the previous one through parameter mapping.

Step 5: Test and Publish

  1. Test with real input data to evaluate the AI output quality
  2. Adjust your prompt based on the results
  3. Publish when satisfied

Best Practices

  • Be specific in your prompts — vague instructions produce vague results
  • Ask for structured output (JSON) when you need to use the result in other nodes
  • Include examples in your prompt for better accuracy
  • Test with varied inputs to ensure the AI handles edge cases
  • Use Router after AI Chat Model nodes to handle unexpected or empty responses