Choosing the Right Action Type

Social Intents offers nine distinct AI Action types, each designed for a different automation scenario. Choosing the right type is the most important decision you make when creating an action - it determines how and when the action fires, what configuration options are available, and how the visitor experiences the result. This guide compares every action type side by side so you can confidently select the right one for your use case.

Quick Comparison Table

Use this table to quickly identify the action type that fits your scenario. Each type is explained in detail in the sections below.

Understanding Trigger Mechanisms

Before diving into individual action types, it is important to understand the three different trigger mechanisms. How an action fires is just as important as what it does.

AI-Triggered Actions

Action types Call API Request, Escalate Chat to Human, Show Button or Iframe Only, and End the Chat are all triggered by the AI engine. During each conversation turn, the AI engine reads the visitor's message, evaluates it against all available action trigger descriptions (the When to use field), and decides whether to fire an action. This means the action only runs when the visitor's intent matches your trigger description.

With AI-triggered actions, you have full control over the trigger logic through natural language. You describe when the action should fire, and the AI engine handles the matching. This is the most flexible and common trigger mechanism.

Auto-Triggered Actions

Action types Auto Trigger on Chat Start, Auto Trigger on Chat Escalation, Auto Trigger on Chat End, and Auto Trigger on Every Message fire automatically based on conversation lifecycle events. They do not rely on the AI engine to evaluate intent - they simply run when the specified event occurs. Auto-triggered actions are event-driven automations that happen behind the scenes.

These actions always require an API endpoint because they execute server-side operations. The visitor may or may not see a response - it depends on whether you set a triggered response message.

Agent-Triggered Actions

The Agent Manually Triggered type fires only when a human agent clicks the action button during a live chat. This is not automated - the agent decides when and whether to run it. This is useful for workflows that require human judgment, like applying a discount, escalating to a manager, or confirming account details.

Detailed Action Type Breakdown

1. Call API Request

This is the most versatile and commonly used action type. It sends an HTTP request to an external API endpoint when the AI detects the visitor's intent matches your trigger description.

When to use it:

• Creating leads or contacts in HubSpot, Salesforce, or Dynamics 365
• Creating support tickets in Jira, Zendesk, Freshdesk, or your custom system
• Looking up order status, account balances, or product availability
• Triggering Zapier or Make webhooks for multi-step automations
• Sending data to your own internal APIs or databases
• Querying a knowledge base or product catalog

Configuration fields:

• HTTP Method - GET, POST, PUT, or DELETE
• Endpoint URL - The full URL of the API endpoint
• Collect data inputs - Parameters the AI needs to collect from the visitor (name, type, description, required flag)
• Request JSON template - Optional JSON body for POST/PUT requests
• Request headers - Authentication headers, content type, API keys
• Expected response fields - Fields to extract from the API response and include in the chat

Example use case: A visitor asks "What's the status of my order #12345?" The AI triggers the API action, sends a GET request to your order API with the order number, and displays the status from the response in the chat.

For a complete guide, see Custom API Actions.

2. Escalate Chat to Human

This action routes the conversation from the AI chatbot to a specific channel or space in your connected messaging platform. It is built for situations where the visitor needs human help, but you want to control which team or channel handles the request.

When to use it:

• Routing billing questions to a #billing channel in Slack
• Sending technical support requests to a specific Teams channel
• Directing sales inquiries to a Google Chat space for your sales team
• Escalating VIP or high-priority requests to a dedicated Webex room

Configuration fields:

• Target Channel / Space - The specific channel, team, or space to route the chat to. The dropdown auto-populates with available channels from your connected platform.
• Routing Instructions - Context message sent to agents in the target channel explaining why this chat was routed to them.
• Require visitor confirmation - When checked, the visitor sees a confirmation button before the chat is escalated.
• Confirmation Button Label - Text on the confirmation button (e.g., "Connect me to a specialist").

This action type does not require an API endpoint - the routing is handled internally by Social Intents.

For a complete guide, see Escalation Routing.

3. Show Button or Iframe Only

This action displays visual elements in the chat - clickable buttons, links, or embedded iframes. It does not make an API call. Use it when you want to present the visitor with an interactive element rather than text alone.

When to use it:

• Embedding a Calendly or Cal.com scheduling widget
• Displaying a payment form or donation widget
• Showing buttons that link to pricing, documentation, or signup pages
• Embedding videos, surveys, or any iframe-compatible content

Configuration fields:

• Show action button in response - Enables the button display
• Button Label - Text shown on the button
• Button URL - The link the button points to, or the URL to embed as an iframe
• Embed as iframe instead of button - When checked, the URL is rendered inside the chat as an embedded iframe rather than a clickable button

For a complete guide, see Showing Buttons, Links & Embedded Content.

4. Auto Trigger on Chat Start

This action fires automatically at the beginning of every chat session. No AI intent matching is needed - it runs as soon as the visitor initiates a chat.

When to use it:

• Calling an API to identify returning visitors by their email or IP
• Logging the start of every chat to an external analytics system
• Pre-loading visitor data from your CRM before the conversation begins
• Triggering a notification webhook when a new chat starts

Key consideration: Because this fires on every single chat, make sure the API endpoint is fast and reliable. A slow or failing endpoint will delay the chat experience for every visitor.

5. Auto Trigger on Chat Escalation

This action fires automatically when the chatbot escalates to a human agent - regardless of the reason for escalation. It runs alongside the escalation event, not instead of it.

When to use it:

• Creating a CRM lead automatically when escalation happens
• Sending a notification to a monitoring tool when the AI hands off to a human
• Logging escalation events for reporting and analysis
• Triggering a Zapier automation on escalation

Key consideration: This is different from the Escalate Chat to Human action type. That type uses AI intent matching to route chats to specific channels. This type fires automatically on any escalation event and is used for side-effect operations like API calls and logging.

6. Auto Trigger on Chat End

This action fires automatically when a chat session ends - when the visitor closes the chat window, navigates away, or the session times out.

When to use it:

• Sending post-chat satisfaction surveys
• Syncing the conversation transcript to your CRM or data warehouse
• Triggering follow-up email sequences through your marketing automation tool
• Logging completed conversations to an analytics platform

7. Auto Trigger on Every Message

This action fires on every single message in the conversation - both visitor messages and bot responses. This is the most intensive action type and should be used sparingly.

When to use it:

• Real-time message logging to an external system
• Continuous sentiment analysis on every visitor message
• Live data enrichment or context injection on each turn
• Streaming conversation data to a monitoring dashboard

8. Agent Manually Triggered

This action is never triggered by the AI or by conversation events. It only fires when a human agent manually clicks the action button from the chat interface during a live conversation.

When to use it:

• Agent-assisted discount offers - the agent clicks to generate a coupon code
• Account lookup - the agent queries a CRM or billing system by customer ID
• Manual ticket creation - the agent decides when to create a support ticket
• Any workflow that requires human judgment before execution

Key consideration: This action requires an agent to be present and actively managing the chat. It does not work in AI-only chatbot mode where no humans are monitoring conversations.

9. End the Chat

This action gracefully closes the chat session. The AI detects when the conversation appears complete - the visitor says "thanks, that's all I needed" or "goodbye" - and triggers the end-chat action. You can include a farewell message.

When to use it:

• Automatically ending resolved conversations with a professional closing message
• Preventing stale chat sessions from remaining open indefinitely
• Delivering a farewell message with next steps, links, or contact information

Configuration: Use the Automatically close the chat after this action runs checkbox along with a closing message. You do not need an API endpoint.

Decision Flowchart

Use this flowchart to determine which action type you need:

1. Does the action need to call an external API?
   • Sim → Do you want the AI to trigger it based on visitor intent?
     • Yes → Use Call API Request
     • No, on chat start → Use Auto Trigger on Chat Start
     • No, on escalation → Use Auto Trigger on Chat Escalation
     • No, on chat end → Use Auto Trigger on Chat End
     • No, on every message → Use Auto Trigger on Every Message
     • No, agent decides → Use Agent Manually Triggered
   • Não → What do you want to happen?
     • Route to a specific team/channel → Use Escalate Chat to Human
     • Show a button, link, or iframe → Use Show Button or Iframe Only
     • Close the chat → Use End the Chat

Performance Considerations

When choosing action types and configuring multiple actions, keep these performance guidelines in mind:

• Auto Trigger on Every Message is expensive - It fires on every single message. Only use it when per-message processing is genuinely required (real-time logging, sentiment analysis). For most use cases, Auto Trigger on Chat End is a better choice.
• Keep trigger descriptions distinct - If two actions have overlapping trigger descriptions, the AI may struggle to choose the right one. Make each When to use description clearly different.
• Fewer actions means faster matching - The AI evaluates every action's trigger for every message. Ten focused actions will perform better than thirty vague ones.
• API endpoint response time matters - Slow API endpoints create delays in the chat experience. Your endpoints should respond in under 2 seconds for the best visitor experience.
• Use trigger-once when appropriate - Prevent the same action from firing multiple times in a conversation by enabling Trigger only once per session.

Perguntas frequentes

Can I change an action's type after creating it?

Yes. You can edit an existing action and change its type from the dropdown. The form will update to show the configuration fields relevant to the new type. Any fields that do not apply to the new type will be hidden but their data is preserved in case you switch back.

What happens if two actions have similar triggers?

The AI engine evaluates all action triggers and picks the best match. If two triggers are very similar, the AI may occasionally pick the wrong one. To avoid this, write clearly distinct trigger descriptions. For example, instead of two actions both triggered by "when someone asks about billing," make one "when the visitor wants to pay an invoice or update payment method" and the other "when the visitor has a question about charges or pricing."

Do auto-triggered actions slow down the chat?

Auto-triggered actions run asynchronously in the background by default. For most action types, the visitor does not wait for the API response. The exception is Auto Trigger on Every Message, which adds processing for every message and can create noticeable delays if the endpoint is slow.

Can I test an action without making it live?

You can test actions by opening your chat widget in preview mode or on a staging page. There is no formal "draft" mode for individual actions, but you can write a very specific trigger description that only matches a test phrase you know to use.

Which action type should I use for Calendly?

Use Show Button or Iframe Only with the Embed as iframe option checked and your Calendly URL in the Button URL field. This embeds the Calendly scheduling widget directly in the chat. See Booking Meetings with Calendly for the complete setup guide.

Which action type should I use for CRM lead capture?

Use Call API Request with the HTTP method set to POST and the endpoint pointing to your CRM's contact/lead creation API (HubSpot, Salesforce, or Dynamics 365). Define the required parameters (email, name, company) as data inputs so the AI collects them from the conversation. See Capturing Leads to HubSpot, Salesforce & Dynamics 365 for detailed setup.