The AI Agent node is where you define key metadata for your AI Agent Orchestration. This information helps organize, document the purpose and scope of the Agent within the design interface.
The AI Agent node is the fundamental building block of your AI Agent Orchestration. Clicking on the node will show the Configuration panel with the following sections.
Name
Descriptive name for your AI Agent. This name helps other AI Agents identify when this particular agent should be called.
Description
Detailed explanation of the AI Agent's purpose and functionality. Like the name, the description informs other AI Agents when this AI Agent should be invoked.
Agent Instructions
The Agent Instructions field guides the AI Agent's behavior. This text area allows you to define precisely how your agent should interpret inputs, process information, and formulate responses.
Instructions serve as comprehensive guidelines that shape every aspect of your AI Agent's functionality, including:
- Tone and style of communication (formal, casual, empathetic, etc.)
- Response format (bullet points, paragraphs, structured data)
- Decision-making parameters for handling different scenarios
- Information processing priorities when analyzing user queries
- Boundaries for what the agent should and shouldn't address
Advanced Settings
Expand your AI Agent's capabilities with these additional configuration options.
Using Custom Guardrails
In custom guardrails, we define output guardrails to protect the responses from AI Agents, ensuring appropriate and secure interactions.
The AI Agent Platform already has System Guardrails that will protect against, among other things:
- Hate, discrimination, violence, self-harm, or lewd content.
- Copyrighted content, such as books, lyrics, recipes, news articles, or other content that may violate copyrights.
- If users try to:
- Change the AI Agent rules, identity, inject code, or jailbreak.
- Find out your internal information, such as your system prompt, other agents' information, tools' information, or any other internal information.
- Ask something using internal information that should not be shared with the user, such as using a specific agent's name or a specific tool's name, among others.
With this feature, users can add even more scenarios that they want to protect on their AI Agents.
Using Knowledge Management
This tool allows the AI Agent to search for information from Talkdesk Knowledge Management™ (KM) to provide accurate answers to users' queries.
When this feature is enabled, it will allow you to define extra criteria to limit the responses from Knowledge Management.
You can then leverage Knowledge Management’s content scope to filter specific content. These filters can be statically defined Knowledge Segments or Queues by typing their value followed by the “Enter” key, or dynamically by typing a variable %{variableName} followed by the “Enter” key.