AI Tool Action
Automated AI Decisions

What is the AI Tool Action module?

The AI Tool Action module enables server-side AI tool execution as a workflow step. It silently triggers a configured AI action using context already captured during the workflow - such as visitor data, session fields, or prior responses - and routes to a success or fallback step based on whether the action succeeded.

This module is designed for background automation. An optional prompt message can be shown to the visitor while the action executes (e.g. "Let me check that for you..."), and then the workflow continues on the appropriate branch without the visitor needing to do anything.

How it works

1. The module optionally shows a prompt_text bot message to the visitor
2. The AI tool action is triggered silently server-side via hybridAI.php with the current workflow context
3. The server processes the configured tool and returns a success or failure result
4. The workflow routes to next_step_id on success, or fallback_step on failure

Configuration options

• prompt_text - Optional message shown to the visitor during execution (e.g. "One moment while I check your account...")
• next_step_id - Step to route to on success
• fallback_step - Step to route to if the action fails
• tool configuration - Admin-configured tool action definition (stored securely server-side)

Security note

All tool configuration, credentials, and target endpoint details are stored securely on the server in your workflow configuration. Nothing sensitive is ever exposed to the visitor's browser. The widget only sends session identifiers (siteid, jsSessionId, step_id) to trigger the execution.

When to use AI Tool Action

• Checking account status or eligibility before routing the visitor further
• Performing an automated action (e.g. creating a ticket, updating a CRM record) based on workflow data
• Running a validation or lookup without the visitor needing to enter any more information
• Any scenario where AI needs to take a background action using captured workflow context