What the AI Agent Action Does
The AI Agent evaluates your instructions and determines:
Which tools to use
The order of execution
What inputs to apply
What actions to complete
This allows one AI Agent action to replace multiple workflow steps such as sending messages, updating records, or generating booking links.
Key Benefits
Autonomous execution using selected tools
Faster workflow setup with fewer steps
Template-based starting points
Custom instruction control
Optional memory for repeat interactions
Execution visibility via logs
Optional structured output for later workflow steps
How to Add an AI Agent Action
Go to Automation > Workflows
Click Create Workflow or open an existing workflow
Open the workflow options menu and click Edit workflow
In the builder, click + Add
Select AI Agent
Configure the action
Click Save action
Adding a Trigger
The workflow must include a trigger before the AI Agent runs.
Common triggers include:
Form Submitted
Contact Created
Appointment Booked
Tag Added
Pipeline Stage Changed
Using Templates
Templates prefill the Instructions and Tools fields for common use cases.
If a template requires a specific trigger, it will be unavailable until that trigger is added.
Available templates include:
Build Your Own
Form Lead Follow Up
No Show Appointment Recovery
Facebook Lead Nurturing
Stale Deal Nudge Agent
Lead Research & Enrichment
New Appointment: Enrich & Confirm
Quiz Lead Scoring & Routing
Call Transcript Summary & Action Items
Task Creation from Won Deals
Instagram / Facebook Comment-to-DM
Lead Pipeline Tracker
Writing Instructions
Instructions define what the agent does, how it makes decisions, and how it uses tools.
Include:
Role definition
Workflow context
Step-by-step actions
If/then logic
Messaging rules
Constraints
Use Enhance Prompt to improve structure and clarity.
Configuring Models and Tools
The model determines how the AI Agent processes instructions. Tools define what actions the AI Agent is allowed to perform.
Common tools include:
Send SMS
Send email
Generate One Time Booking Link
Create Opportunity
Add Contact Tag
Update Contact Field
Send Internal Notification
Advanced Options
Advanced settings control memory and output behavior.
Conversation Memory: Uses a rolling summary of past runs for the same contact
Output Format: Controls whether the agent returns None, Text, or JSON
Use JSON when later workflow steps need structured data.
Built-in Conversation History
AI Agent can retrieve past conversations automatically when additional context is required.
This includes:
Channel
Message content
Direction
Timestamps
Summary
Execution Logs
Execution Logs show what the AI Agent did during a workflow run.
You can review:
Agent Input
LLM steps
Tool executions
Execution time
Token usage
Final output
Pricing
AI Agent usage may include:
Model token usage
Number of executions
Premium tool usage
Total cost depends on the model selected, tokens consumed, and tools used.
Common Issues and Fixes
Template is unavailable
Add the required trigger, then reopen the template selector.
AI Agent does not use a tool
Ensure the tool is added and clearly referenced in the instructions.
Output cannot be used later
Set Output Format to Text or JSON.
Repeated messaging
Enable Conversation Memory for recurring workflows.
FAQ
Is there a limit on tools?
Yes. You can add up to 10 tools. Internal tools do not count toward this limit.
Which model should I use?
Use GPT-5.2 (Low thinking) for most use cases. Use higher thinking models for complex logic.
Can AI Agent work with any trigger?
Yes. AI Agent works with any trigger. Templates may require specific triggers.
What does Enhance Prompt do?
It restructures instructions into a clearer and more detailed format.
What is the difference between Conversation Memory and Conversation History?
Conversation Memory is a rolling summary of prior runs. Conversation History is retrieved messaging context when needed.
When should I use JSON output?
Use JSON when later workflow steps need structured values.
How do I view execution details?
Open Execution Logs and select the AI Agent run.










