Skip to main content

Overview

AI agents are the “personalities” that handle your phone calls. Each agent has its own instructions, voice, and behavior configuration. This guide explains how to create, configure, and manage agents.

What is an Agent?

An agent is a configured AI assistant that:
  • Handles phone conversations automatically
  • Follows specific instructions you provide
  • Uses a selected voice model
  • Operates in a specified language
  • Can extract structured data from conversations

Creating an Agent

Using the Developer Portal

  1. Navigate to the Agents section
  2. Click Create New Agent
  3. Fill in the agent configuration:
    • Name: Descriptive name for your agent
    • Instruction: The prompt that defines agent behavior
    • Voice: Select a voice model
    • Language: Choose the conversation language
  4. Click Save

Using the API

You can create agents programmatically using the API (if available) or use temporary agents by providing an instruction when making calls.

Agent Instructions

The instruction (or prompt) is the most important part of your agent. It defines:
  • How the agent behaves
  • What the agent says
  • How the agent responds to different situations
  • What information to collect

Writing Effective Instructions

Best Practices

  1. Be Specific: Clearly define the agent’s role and responsibilities
  2. Set Boundaries: Specify what the agent should and shouldn’t do
  3. Provide Examples: Include example responses for common scenarios
  4. Define Goals: State what the agent should accomplish
  5. Handle Edge Cases: Include instructions for unexpected situations

Agent Configuration

Voice Selection

Choose a voice that matches your brand and use case:
  • Professional: For business and customer service
  • Friendly: For sales and marketing
  • Calm: For support and troubleshooting
  • Energetic: For promotions and announcements

Language Settings

Select the primary language for conversations. The agent will:
  • Speak in the selected language
  • Understand responses in that language
  • Adapt to regional variations when possible

Temporary Agents

You can create temporary agents on-the-fly by providing an instruction when making a call:
Temporary agents are useful for one-off calls or testing, but persistent agents are recommended for production use.

Structured Data Extraction

Agents can extract structured data from conversations using output_parameters:

Example: Appointment Scheduling

The agent will extract these fields during the conversation, and they’ll be included in the webhook payload when the call completes.

Output Parameters Template

Define the structure you want to extract:

Agent Management

Updating Agents

You can update agent configurations through:
  • Developer Portal: Edit agent settings in the web interface
  • API: Update agents programmatically (if supported)

Testing Agents

Before using an agent in production:
  1. Test with your own phone number
  2. Try different conversation scenarios
  3. Verify the agent follows instructions correctly
  4. Check that data extraction works as expected

Agent Analytics

Monitor agent performance:
  • Call success rate: Percentage of successful calls
  • Average duration: Average call length
  • Goal completion: How often agents achieve their objectives
  • Customer satisfaction: Feedback and ratings

Common Use Cases

Customer Service

Handle customer inquiries, support requests, and troubleshooting

Appointment Scheduling

Schedule and confirm appointments automatically

Lead Qualification

Qualify leads and collect contact information

Order Confirmation

Confirm orders and provide delivery updates

Survey Collection

Conduct surveys and collect feedback

Reminders

Send reminders for appointments and deadlines

Best Practices

  1. Start Simple: Begin with basic instructions and refine over time
  2. Test Thoroughly: Test agents with various scenarios before production
  3. Iterate: Continuously improve instructions based on call results
  4. Monitor Performance: Track metrics and adjust as needed
  5. Handle Errors: Include error handling in your instructions
  6. Set Expectations: Clearly communicate what the agent can and cannot do

Troubleshooting

  • Review and refine your instructions
  • Be more specific about desired behavior
  • Add examples of correct responses
  • Test with different phrasings
  • Add explicit instructions about response length
  • Specify when to ask follow-up questions
  • Define when to end the conversation
  • Verify output_parameters structure is correct
  • Ensure instructions mention what data to collect
  • Test with clear examples
  • Check webhook payload for extracted data

Examples

Customer Support Agent

Appointment Scheduler

Support

For agent-related questions: