AI Assistants
AI Assistants enable automated call handling using artificial intelligence. This document explains how to configure and use AI Assistants in OPBX.
What are AI Assistants?
AI Assistants are virtual agents that can:
- Answer and handle incoming calls automatically
- Understand natural language using AI
- Route calls to appropriate departments
- Answer common questions
- Take messages
- Provide 24/7 availability
Creating an AI Assistant
Step 1: Open Create Dialog
- Navigate to AI Assistants in the main menu
- Click Create AI Assistant
Step 2: Configure Basic Settings
Fill in the required fields:
| Field | Description | Required |
|---|---|---|
| Name | Display name for the AI Assistant (e.g., "Customer Service Bot") | ✅ Yes |
| AI Service Provider | Select from available providers | ✅ Yes |
Step 3: Select Provider Protocol
Providers are grouped by protocol type:
- SIP Providers - Traditional telephony integration
- WebSocket Providers - Modern web-based integration
Select the appropriate provider for your use case.
Step 4: Configure Provider-Specific Settings
After selecting a provider, additional configuration fields will appear based on the provider's requirements. These may include:
- API Key - Authentication token for the AI service
- Bot ID - Identifier for the specific bot/agent
- Auth Token - Additional authentication credentials
- Assistant ID - Provider-specific assistant identifier
- Workspace/Project ID - Organization context
- Phone Number - Number assigned to the AI (for SIP providers)
Configuration fields vary by provider. Required fields are marked with an asterisk (*).
Step 5: Save and Activate
- Review all configuration fields
- Click Save
- The AI Assistant is created with "Active" status by default
Step 6: Test the AI Assistant
Before deploying to production:
- Review the AI Assistant details
- Test call handling if a test interface is available
- Verify configuration with the AI provider
- Monitor initial call interactions
Managing AI Assistants
Viewing AI Assistants
The AI Assistants page displays:
- Name - Assistant display name
- Provider - Service provider (e.g., Twilio, OpenAI)
- Protocol - SIP or WebSocket
- Status - Active or Inactive
- Usage - Number of extensions using this assistant
Editing an AI Assistant
- Find the AI Assistant in the list
- Click Edit (pencil icon)
- Modify fields as needed:
- Name
- Provider (note: changing provider resets configuration)
- Configuration fields
- Click Save
Changing the provider will reset all configuration fields. You will need to re-enter provider-specific settings.
Activating/Deactivating
Toggle the status of an AI Assistant:
- Click the Status toggle in the AI Assistant row
- Or edit the assistant and change the status
Inactive AI Assistants cannot receive calls.
Deleting an AI Assistant
An AI Assistant cannot be deleted if it's currently in use by one or more extensions. Remove all associations first.
To delete:
- Ensure no extensions are using the AI Assistant
- Click Delete (trash icon)
- Confirm deletion
Routing Calls to AI Assistants
Via DID Number
Route incoming calls directly to an AI Assistant:
- Go to DID Numbers
- Select the phone number to configure
- Set Routing Type to "AI Assistant"
- Select your AI Assistant from the dropdown
- Save changes
Via Extension
Create an extension that routes to the AI Assistant:
- Go to Extensions
- Click Create Extension
- Select AI Assistant as the Extension Type
- Select your AI Assistant
- Assign an extension number
- Save
Now callers can dial the extension number to reach the AI.
Via IVR Menu
Add AI Assistant as an IVR menu option:
- Go to IVR Menus
- Select or create an IVR menu
- Add a menu option (key press)
- Set Destination Type to "AI Assistant"
- Select your AI Assistant
- Save
AI Load Balancers
For high-volume scenarios, use AI Load Balancers to distribute calls across multiple AI Assistants.
What are AI Load Balancers?
AI Load Balancers:
- Distribute calls across multiple AI Assistants
- Prevent overloading a single AI
- Provide redundancy
- Support different distribution algorithms
Creating an AI Load Balancer
- Navigate to AI Load Balancers
- Click Create Load Balancer
- Enter a Name
- Select Distribution Algorithm
- Add AI Assistants as members
- Configure fallback options
- Save
Distribution Algorithms
| Algorithm | Description |
|---|---|
| Round Robin | Distribute evenly in rotation |
| Least Connections | Route to AI with fewest active calls |
| Weighted | Distribute based on configured weights |
Fallback Configuration
Set what happens when all AIs are unavailable:
- Fallback AI - Route to backup AI
- Extension - Route to human agent
- Voicemail - Take a message
Use Cases
After-Hours Support
Setup:
- Main DID → Business Hours
- Open → Ring Group (Support Team)
- Closed → AI Assistant
Benefits:
- 24/7 availability
- Handle common questions
- Take messages for follow-up
High-Volume Handling
Setup:
- Main DID → AI Load Balancer (3 AI Assistants)
- Overflow → Ring Group (Human agents)
Benefits:
- Handle peak call times
- Reduce wait times
- Scale without adding staff
First-Line Support
Setup:
- Main DID → AI Assistant
- AI can transfer to:
- Sales (Press 1)
- Support (Press 2)
- Billing (Press 3)
Benefits:
- Immediate response
- Automatic routing
- Reduce human workload
Best Practices
Configuration
- Clear expectations - Let callers know they're speaking with AI
- Easy escalation - Always provide option to reach human
- Test thoroughly - Verify AI responses before deploying
- Monitor performance - Review call logs regularly
Training
- Document common questions - Ensure AI can answer them
- Update regularly - Add new information as needed
- Review failures - Identify where AI struggles
- Iterate - Continuously improve responses
Security
- Protect API keys - Keep AI provider credentials secure
- Monitor usage - Watch for unusual activity
- Set limits - Configure rate limits if available
- Audit access - Log who configures AI settings
Troubleshooting
AI Not Responding
Problem: Calls reach AI but no response.
Solutions:
- Check AI provider status
- Verify API credentials
- Review provider configuration
- Check phone number assignment
Poor Quality Responses
Problem: AI gives incorrect or unhelpful responses.
Solutions:
- Review AI training/prompts
- Update knowledge base
- Adjust AI model settings
- Consider different provider
Calls Not Reaching AI
Problem: Calls bypass AI routing.
Solutions:
- Verify DID routing configuration
- Check extension type is "AI Assistant"
- Review IVR menu destinations
- Ensure AI Assistant is active
Limitations
Current Limitations
- AI Assistants cannot make outbound calls
- Complex multi-turn conversations may have limitations
- Provider-specific feature limitations apply
- Requires internet connectivity
Next Steps
- Conference Rooms - Set up multi-party conferences
- Call Recording - Record AI-handled calls
- Reporting - Monitor AI call analytics
Related Documentation:
- IVR Menus - Route to AI via menus
- Business Hours - Time-based AI routing
- DID Numbers - Direct AI routing