Behavior Settings
After completing the Basic Information setup, proceed to the Behavior Settings tab. This section defines how your AI agent interacts with users — its tone, purpose, communication style, and example prompts that shape its overall behavior.
Guardrail
Select a Guardrail if you want to enforce specific rules and boundaries on the AI agent. Guardrails prevent the AI from performing unwanted or unsafe actions— such as disclosing confidential data, generating inappropriate content, or violating company communication protocols, maintaining reliability and compliance.
When creating or editing your AI agent, you can select from available guardrails to automatically apply pre-set safety filters.
Available Guardrails:
NVIDIA NEMO – A standard safety and compliance framework that restricts unsafe content generation and enforces ethical AI usage. Recommended for production or customer-facing bots to maintain reliability and trust.
GuardRail Test Hung – A testing configuration for internal experiments or sandbox environments. Recommended only for non-production or trial environments.
Example Use: For a Lead Qualifier AI Agent, select NVIDIA NEMO to ensure all customer interactions remain safe, professional, and compliant with privacy and communication standards.
If no pre-built guardrail is needed, you can leave this field blank for basic configurations.
Personality and Role
Define the AI’s character and role in one or two sentences. This determines how the AI behaves, greets users, and delivers information.
Example Entry: “You are a friendly and professional virtual assistant who acts as a Lead Qualifier for an SME. You help collect customer details such as name, company, requirements, and budget, and guide them to the appropriate sales contact.”
Core Task
Describe the main tasks the AI will perform during interaction. This helps the model stay focused and consistent in its responses.
Example Entry: “Your tasks include gathering lead information, confirming contact details, qualifying potential clients as Hot, Warm, or Cold, and summarizing key inputs for the sales team.”
4. Preload Information
Use this field to give the AI key background data or reference details that it should always know before starting a conversation.
Example Entry: “Our company offers digital marketing, branding, and content management services. Leads are classified based on urgency, budget, and scope. Qualified leads are forwarded to the sales department.”
5. Tone and Style
Define how your AI communicates — its personality and phrasing style.
Example Entry: “Maintain a friendly, clear, and conversational tone. Always greet politely, use short and structured replies, and confirm information before ending the conversation.”
6. Response Length
Specify the average length of each AI reply to ensure clarity and consistency.
Example Entry: “Within 100 words.”
7. List of Banned Words
Identify words or phrases the AI should avoid in responses. This ensures professional and compliant communication.
Example Entry: “‘guaranteed’, ‘confidential’, ‘personal opinion’, ‘refund’, or slang expressions.”
8. Other Requirements (Optional)
Add any custom rules or behavioral notes your AI must follow. If you have any other requirement, click ‘+ Add requirements’ and type your note, then click Post.
Example Entry: “Always confirm user contact details before closing a conversation.”
9. Capabilities
This section allows you to assign workflow-based functions to enhance your AI’s operational power. Functions enable the AI to perform real-world actions automatically.
Examples of Available Functions:
(AI Booking) Send Email – Compose and send an email when a user requests communication.
After selecting a function, click Apply to Assistant to link it with your AI.
Last updated