Skip to main content
Action steps define what happens after a trigger fires. They can be grouped into three main categories: AI steps, core logic steps, and integrated app steps.

​
1. AI Action Steps

Use AI models or Attention’s intelligence to generate outputs or fetch insights.
  • Ask Attention: summarize conversations, extract insights, or build follow-up emails
  • Attention Conversation utilities: list conversations, fetch conversations, or query by ID
  • Ask OpenAI (ChatGPT): use OpenAI’s models for general-purpose tasks
  • Ask Anthropic (Claude): use Anthropic’s models for extended reasoning
Example use cases:
  • Generate a discovery call recap
  • Draft a follow-up email template
  • Extract key metrics across multiple calls

​
2. Core Action Steps

Steps that manage logic, routing, and data handling inside the workflow. These include:
  • Routers: branch execution paths with if/else logic
  • Loops: iterate over lists of items (e.g., conversations or rows)
  • Storage: collect and aggregate outputs across steps
  • Text Helpers: transform or clean text values
  • HTTP Requests: connect to any API by sending a request
  • Custom Code Step: run Node.js & TypeScript code
    • To include data from previous steps, add pairs of keys and values as inputs in the Code Step configuration
Example use cases:
  • Only continue if conversation label = “Discovery Call”
  • Loop through all sales calls this week and append summaries to storage
  • Send data via API

​
3. Integrated App Steps

These steps let you act directly in other SaaS tools. Supported categories include:
  • G Suite: Gmail, Google Sheets, Google Docs
  • CRMs: Salesforce, HubSpot, Pipedrive, Zoho
  • Productivity apps: Asana, Airtable, Notion, Linear
  • Data platforms: Snowflake and others
Example use cases:
  • Send Slack messages with meeting summaries
  • Update Salesforce account plans automatically
  • Log call recaps into a Google Doc