AI Agent (Persona)
Learn how to create and configure your AI agent to handle customer conversations automatically.
Your AI Agent (also called Persona) is your virtual customer service representative that automatically responds to customer messages across all your connected channels. Configure its personality, behavior, and conversation style to match your brand.
What is an AI Agent?
An AI Agent is an intelligent assistant that:
- Responds automatically to customer messages
- Uses your Knowledge Base to provide accurate information
- References your FAQs to answer common questions
- Maintains conversation context throughout interactions
- Handles multiple conversations simultaneously
- Can escalate to human agents when needed
Creating Your AI Agent
Step 1: Access AI Personas
- Navigate to your Cloodot dashboard
- Go to Intelligence → AI Personas (or AI → Personas)
- Select or create a persona to configure
Step 2: Configure Your Agent
Your AI Agent configuration is organized into four tabs:
- Identity: Basic settings and behavior
- Instructions: How your AI should work
- Skills: Installed and default actions
- Advanced Settings: Fine-tuning and fallback options
Let's explore each tab in detail.
Configuring Your Agent
Tab 1: Identity
The Identity tab contains all basic settings for your AI agent.
Basic Information
- Nickname: A friendly name for your AI agent (e.g., "Support Bot", "Jordan")
- Description: Brief description of what your agent does and its purpose
- Greeting: The message sent when a customer says "hi" or starts a conversation
:::tip Your greeting is the first impression customers get. Make it welcoming and clear about what your agent can help with. :::
Preferred Languages
Configure which languages your agent will respond in:
- Available Languages: English, Arabic, Hindi, Urdu, Spanish, French, German, Malayalam
- Selection: Use checkbox cards to select preferred languages
- Understanding: Your AI agent can understand any language, but you can limit the languages in which the agent will respond
:::info Even if you limit response languages, your AI can still understand messages in other languages. It will respond in one of your selected preferred languages. :::
Follow-up Settings
Configure automatic follow-up messages:
- Purpose: If the Agent does not receive a response within the selected time, an automatic follow-up message will be sent
- Options:
- 1 minute
- 10 minutes
- 1 hour
- Custom (set your own time interval)
:::tip Follow-up messages help re-engage customers who haven't responded, improving conversation completion rates. :::
Response Length
Control how detailed your agent's responses are:
- Concise: Short, brief answers
- Medium: Balanced detail and brevity
- Long: Comprehensive, detailed responses
Personality
Choose your agent's conversation style:
- Professional: Formal and business-like
- Friendly: Warm and approachable
- Casual: Relaxed and conversational
Handover Conditions
Configure when your AI should hand over to a human agent:
- Custom Conditions: Set different conditions based on custom fields
- Automatic Handover: AI will automatically transfer to a human agent when conditions are met
- Flexible Rules: Create multiple handover conditions for different scenarios
:::warning Handover conditions are important for ensuring complex issues reach human agents. Configure them carefully based on your business needs. :::
Tab 2: Instructions
The Instructions tab is where you define how your AI should work and behave.
Writing Instructions
In this tab, you can:
- Define Behavior: Write detailed instructions on how your AI should handle conversations
- Set Guidelines: Specify rules, tone, and approach for different situations
- Create Instruction Blocks: Organize instructions into blocks for better structure
Instruction Blocks
Instruction blocks help organize your instructions:
- Add Blocks: Create separate instruction blocks for different scenarios
- Structure: Each block can contain specific guidelines or behaviors
- Quick Replies: Inside instruction blocks, use the "Insert Quick Reply" button to generate buttons in the AI's response text
Quick Replies in Instructions
When writing instructions:
- Place your cursor where you want a quick reply button
- Click "Insert Quick Reply +" button
- The button will be added to the response text
- Customers can click these buttons to continue the conversation flow
:::tip Quick replies in instructions help guide customers through conversation flows and make interactions more interactive. :::
Tab 3: Skills
The Skills tab displays all available actions your AI can perform.
Installed Skills
- Default Actions: Built-in skills available to all agents
- Installed Actions: Skill Sets you've installed and configured
- Skill Management: View, configure, and manage all available skills
Using Skills
Skills enable your AI to:
- Perform specific actions (check orders, book appointments, etc.)
- Integrate with external services
- Execute custom workflows
- Handle specialized tasks
:::info To add more skills, install Skill Sets from the Intelligence → Skill Sets section. :::
Tab 4: Advanced Settings
The Advanced Settings tab allows you to fine-tune behavior and configure fallback options.
Temperature Adjustment
Control your AI's creativity and consistency:
- Range: Adjustable slider or input (typically 0.0-1.0)
- Lower Values: More consistent, predictable responses
- Higher Values: More creative, varied responses
- Default: Usually set to 0.7 for balanced behavior
Fallback Message
Configure what happens when AI fails to respond:
- Fallback Message: Message sent when the AI cannot generate a response
- Purpose: Ensures customers always receive a response, even if AI encounters issues
- Customization: Write your own fallback message or use the default
:::warning A good fallback message is essential. It should apologize for the issue and offer to connect the customer with a human agent. :::
Fine-tuning Options
Additional advanced settings may include:
- Response Formatting: Control how responses are structured
- Error Handling: Configure how to handle errors
- Performance Settings: Optimize response speed and quality
Saving and Deploying Your Agent
Saving Changes
- After configuring all tabs, click "Save" button
- Your changes will be saved to the current version
- You can continue editing or deploy when ready
Viewing Versions
- View All Versions: Click "View All Versions" to see version history
- Version Management: Switch between versions, compare changes, or roll back
- Version Control: Each save creates a new version for tracking changes
Testing Your Agent
- Click the "Play" button to open the playground
- Test your agent's responses with sample conversations
- Verify instructions, skills, and behavior work as expected
- Make adjustments and test again before deploying
Deploying Your Agent
- Once satisfied with configuration, deploy your agent
- The deployed version will be active and handle customer conversations
- You can update and redeploy at any time
Using Your Agent
Automatic Responses
Once deployed, your agent will:
- Automatically respond to new customer messages
- Use your Knowledge Base to find relevant information
- Reference your FAQs for common questions
- Follow your instructions and guidelines
- Maintain context throughout conversations
Conversation Handling
Your agent can:
- Handle multiple conversations simultaneously
- Remember context within a conversation
- Use preferred languages for responses
- Send follow-up messages when configured
- Escalate to human agents based on handover conditions
- Use installed skills to perform specific actions
Monitoring Performance
Track your agent's performance:
- View conversation analytics
- Monitor response quality
- Check escalation rates
- Review customer satisfaction
- Analyze skill usage
Agent Versions
Version Management
Your agent uses versioning to track changes:
- Automatic Versions: Each save creates a new version
- Version History: View all previous versions and their changes
- Rollback: Revert to any previous version if needed
- Testing: Test new versions before deploying
Working with Versions
- Create New Version: Make changes and save to create a new version
- Test Version: Use the playground to test before deploying
- Deploy Version: Activate a version to handle customer conversations
- Compare Versions: See what changed between versions
- Rollback: Switch back to a previous version if needed
Best Practices
Identity Tab Configuration
- Clear Nickname: Choose a memorable, brand-appropriate name
- Helpful Description: Clearly explain what your agent does
- Engaging Greeting: Make the first message welcoming and informative
- Language Selection: Select languages your customers actually use
- Appropriate Follow-up: Set follow-up timing based on your business needs
- Right Response Length: Match response length to your use case
- Consistent Personality: Choose a personality that matches your brand
- Clear Handover Rules: Define specific conditions for human handover
Instructions Tab
- Be Specific: Write clear, detailed instructions
- Use Instruction Blocks: Organize instructions into logical blocks
- Add Quick Replies: Use quick replies to guide conversation flows
- Regular Updates: Update instructions based on common questions and issues
- Test Instructions: Use playground to verify instructions work as expected
Skills Tab
- Install Relevant Skills: Only install Skill Sets you actually need
- Configure Properly: Ensure all skills are properly configured
- Keep Updated: Update installed Skill Sets regularly
Advanced Settings
- Temperature Balance: Find the right balance between consistency and creativity
- Effective Fallback: Write a helpful fallback message that offers human assistance
- Regular Review: Review and adjust advanced settings based on performance
General Tips
- Test Thoroughly: Always test in playground before deploying
- Monitor Performance: Regularly review conversation analytics
- Iterate: Continuously improve based on real customer interactions
- Use Knowledge Base: Complement instructions with comprehensive Knowledge Base
- Maintain FAQs: Keep FAQs updated to support your agent
Troubleshooting
Agent Not Responding
Problem: Agent isn't responding to messages
Solutions:
- Verify the agent is deployed
- Check that a version is active
- Ensure the agent is assigned to conversations
- Check channel connections
Poor Response Quality
Problem: Responses are inaccurate or irrelevant
Solutions:
- Add more information to Knowledge Base
- Update FAQs with better answers
- Adjust temperature settings
- Review and improve custom instructions
Agent Escalating Too Often
Problem: Agent hands over to humans too frequently
Solutions:
- Add more information to Knowledge Base
- Create FAQs for common questions
- Adjust escalation criteria in instructions
- Review conversation history to identify gaps
Advanced Topics
Extending Your Agent's Capabilities
- Skill Sets - Add custom actions and integrations to your AI agent
- Creating Custom Skills - Build your own skills to extend functionality
- Skill Schema - Understand the technical structure of skills
- Skill Best Practices - Learn advanced patterns and optimization techniques
Advanced Configuration
- MCP Servers - Connect external tools and services via Model Context Protocol
- Copilot Settings - Configure advanced AI features and integrations
- Knowledge Base - Build comprehensive knowledge repositories
- FAQ Management - Create and manage frequently asked questions
Next Steps
After setting up your AI Agent:
- Build Your Knowledge Base to provide business information
- Create FAQs for common questions
- Explore Skill Sets to add custom capabilities
- Test your agent in the playground before deploying
Ready to enhance your agent? Add knowledge to your Knowledge Base!