Assistant Routing

Routing rules determine which assistant handles each conversation. Create rules to automatically route conversations to the right assistant based on keywords, intent, or other conditions.

How Routing Works

When a conversation starts, the system evaluates routing rules in priority order:

1. Check Active Rules: Only active rules are evaluated
2. Priority Order: Rules with higher priority are checked first
3. Match Conditions: Check if rule conditions match (keywords, intent, channel)
4. Route to Assistant: If matched, route to the specified assistant
5. Fallback: If no rules match, use the default assistant

Routing Rule Components

Each routing rule consists of:

• Rule Name: Descriptive name for the rule
• Priority: Higher numbers are evaluated first (0-100)
• Active Status: Whether the rule is currently active
• Trigger Keywords: Words or phrases that trigger the rule
• Trigger Intent: Intent classification that triggers the rule
• Channels: Which channels this rule applies to (SMS, Chat, Email, Voice)
• Target Assistant: Which assistant to route to when matched

Creating Routing Rules

Rule Priority

Priority determines evaluation order:

• Higher Priority (90-100): Specific, important rules (e.g., "Urgent Support")
• Medium Priority (50-89): General rules (e.g., "Sales Inquiries")
• Lower Priority (0-49): Catch-all rules (e.g., "Default Routing")

Best Practice: Use higher priorities for specific rules, lower priorities for general rules.

Trigger Keywords

Keywords are words or phrases that, when detected in the conversation, trigger the rule.

Examples:

• ["support", "help", "issue"] - Routes support conversations
• ["pricing", "cost", "price"] - Routes pricing inquiries
• ["demo", "trial", "test"] - Routes demo requests

Tips:

• Use lowercase keywords
• Include common variations
• Be specific to avoid false matches
• Test with real conversations

Trigger Intent

Intent-based routing uses AI to classify conversation intent and route accordingly.

Common Intents:

• support - Customer support requests
• sales - Sales inquiries
• billing - Billing questions
• technical - Technical questions

Note: Intent classification requires AI processing and may have a slight delay.

Channel Selection

Specify which channels a rule applies to:

• SMS: Text message conversations
• Chat: Web chat conversations
• Email: Email conversations
• Voice: Phone call conversations

You can select multiple channels or all channels.

Target Assistant

Select which assistant should handle conversations that match this rule. The assistant must exist and be active.

Rule Evaluation Logic

Rules are evaluated as follows:

1. Check if Active: Inactive rules are skipped
2. Check Channel: Rule must apply to the conversation's channel
3. Check Keywords: If keywords are specified, at least one must be found
4. Check Intent: If intent is specified, conversation intent must match
5. Route: If all conditions match, route to target assistant

Important: Rules are evaluated in priority order. The first matching rule wins. Lower priority rules are not evaluated if a higher priority rule matches.

Tenant-Level vs Assistant-Level Rules

Tenant-Level Rules

• Apply to all conversations across all assistants
• Use when you want global routing logic
• Set assistant_id to null/empty

Assistant-Level Rules

• Apply only to conversations for a specific assistant
• Use when you want assistant-specific routing
• Set assistant_id to the assistant's ID

Example: A tenant-level rule might route all "support" conversations to a support assistant, while an assistant-level rule might route specific keywords within that assistant's conversations to a specialized sub-assistant.

Best Practices

1. Start with Default: Begin with a default assistant, add rules as needed
2. Test Rules: Test routing rules with sample conversations
3. Monitor Performance: Review which rules are matching and adjust as needed
4. Use Priorities Wisely: Higher priority for specific rules, lower for general
5. Keep Rules Active: Deactivate rules instead of deleting to preserve history
6. Document Rules: Use clear rule names that explain their purpose

Common Routing Patterns

Pattern 1: Channel-Based Routing

Route different channels to different assistants:

• Voice → Voice-optimized assistant
• Email → Email-optimized assistant
• Chat → Chat-optimized assistant

Pattern 2: Intent-Based Routing

Route based on conversation intent:

• Support intent → Support assistant
• Sales intent → Sales assistant
• Technical intent → Technical assistant

Pattern 3: Keyword-Based Routing

Route based on keywords:

• "pricing", "cost" → Sales assistant
• "help", "support" → Support assistant
• "bug", "error" → Technical assistant

Pattern 4: Hybrid Routing

Combine multiple conditions:

• Support intent + "urgent" keyword → Priority support assistant
• Sales intent + "enterprise" keyword → Enterprise sales assistant

Troubleshooting

Rule Not Matching

• Check if rule is active
• Verify keywords are spelled correctly (lowercase)
• Ensure rule applies to the correct channel
• Check rule priority (higher priority rules may be matching first)

Wrong Assistant Selected

• Review rule priority order
• Check if multiple rules might be matching
• Verify target assistant is active
• Test with sample conversations

Rules Not Working

• Ensure at least one default assistant is set
• Check that target assistants exist and are active
• Verify rule conditions are correct
• Review rule priority values

Next Steps

• Learn about Assistants Overview - Assistant concepts
• Learn about Configuration - Assistant settings
• Learn about Handoff Settings - Configure handoff behavior
• Review Conversations - See routing in action