Using the AI Agent Workflow Action
What is the AI Agent Workflow Action?
The AI Agent is a premium action that can plan and execute multi-step tasks within a workflow. Instead of manually building a sequence of individual actions, you can define a goal in plain language. The AI Agent will then autonomously decide which available tools to use and in what order to achieve that goal. This action is ideal for processes requiring flexible decision-making or complex sequences of events.
Configuring the AI Agent Action
To add and configure an AI Agent, locate it in your workflow builder's action library and drag it into your workflow canvas. The configuration panel contains several key settings.
Action Name
Give the action a descriptive name to help you identify its purpose within the workflow. The default name is "AI Agent."
Template
You can start from a pre-built template or create your own agent. Templates provide a starting configuration of instructions and tools for common scenarios, such as following up on a form submission or recovering a missed appointment. If you select Build Your Own, you will configure the agent from scratch.
Instructions
This is the most critical part of the setup. In the instructions field, clearly describe the task you want the AI Agent to perform. Define the goal, any rules it must follow, and how it should make decisions. Be as specific and detailed as possible for the best results.
An Enhance Prompt button is available to help reformat and improve loosely written instructions. You can undo this enhancement if needed.
Model
Select the underlying AI model that will power the agent's reasoning and execution.
Tools
Here, you select which specific workflow actions the AI Agent is permitted to use. The agent can only utilize the tools you explicitly enable in this section. Available tools include actions for managing contacts, sending communications, creating records, and more.
Conversation Memory
When enabled for a specific contact, this feature provides the agent with a summary of its previous executions on that same workflow step. This gives the agent context about past interactions with that contact.
Output Format
Choose the format for the agent's final output, which can be used in subsequent workflow steps. Options include plain text or structured JSON.
Reviewing Execution and Understanding Costs
Execution Logs
After an AI Agent runs, you can review detailed execution logs. These logs show the agent's step-by-step reasoning, which tools it used, the inputs it provided to those tools, and the total tokens consumed during the run.
Pricing
The AI Agent is a premium feature. Each execution of this action will incur an additional charge based on token usage. You can see the token consumption for each run in the execution logs.