AI Auto-Reply Setup
AI Auto-Reply is an intelligent automation system that generates and sends responses to incoming customer messages automatically using AI. It learns from your knowledge base and past conversations to provide accurate, on-brand responses without requiring manual agent input.
Key Benefits
- Instantly respond to customer messages 24/7
- Reduce response time and improve customer satisfaction
- Use your knowledge base to provide accurate answers
- Learn from past conversations to improve response quality
- Control with conditions and moderator review for safety
- Optional image analysis for visual support questions
How AI Auto-Reply Works
AI Auto-Reply processes incoming messages through multiple stages to ensure quality responses.
Message Received
A customer sends a new message to your chat. The system captures it and begins processing.
Conditions Evaluated
If conditions are configured, the AI checks if the message matches your rules (e.g., "only reply to pricing questions"). Messages that don't match are skipped.
AI Generates Response
The AI uses your knowledge base and conversation context to generate a relevant, on-brand response to the customer message.
Moderator Review (Optional)
If moderators are assigned, the response enters a review queue. They can approve, edit, or reject before it reaches the customer. Without moderators, it goes straight to delivery.
Response Delivered
The approved response is sent to the customer, closing the automation loop.
How it learns: The AI improves over time by learning from your company context (tone, policies, industry), conversation examples you provide, and your knowledge base articles.
Setup Guide
Enable AI Auto-Reply
- Navigate to Automations in the sidebar
- Click on AI Auto-Reply
- Toggle the AI Auto Reply switch to enable the feature
- You'll see a "How it works" explanation and configuration options
Add Conditions (Optional)
Conditions let you filter which incoming messages should get an AI response. Leave empty to reply to all messages.
- Click the Add button in the Conditions section
- Use the Condition Builder to create rules (e.g., "message contains pricing")
- Combine multiple conditions with AND/OR logic
- Click Save Conditions when done
Assign Moderators (Recommended)
Moderators review AI-generated responses before they reach customers. This is especially recommended when you first enable AI Auto-Reply.
- Click Assign in the Moderators section
- Select team members who will review responses
- Moderators will receive responses in a review queue
- You can remove or add moderators anytime
Without moderators, AI responses are sent immediately. Adding moderators introduces a safety layer for quality assurance.
Configure Reply to Images (Optional)
- Toggle Reply to Images to enable image analysis
- A confirmation modal will show token usage implications
- Each image uses approximately 60,000–100,000 tokens
Token Usage Warning: Image analysis is expensive. Only enable if you have moderate image message volume and adequate token budget.
Set Up Context (Important)
Context makes AI responses more accurate and on-brand. Configure company-wide guidelines and provide conversation examples.
- Navigate to the Knowledge Hub section (linked in AI Auto-Reply)
- Add Company Context: industry, tone, policies, company name
- Add Conversation Examples: sample customer messages and your team's responses
- Mark contexts as Active to use them for AI training
Managing Conditions
Conditions allow you to control which messages trigger AI Auto-Reply. They work like filters on your inbox.
Condition Examples
- "message contains price" — reply only to pricing questions
- "message contains refund" — reply to refund-related queries
- "message contains account" — reply to account issues
- "not from returning customer" — reply only to new customers
Condition Tips
- Leave empty to reply to all messages
- Use specific keywords for accuracy
- Combine rules with AND for strict filtering
- Use OR to broaden what triggers replies
Moderators Setup
Moderators are team members who review AI-generated responses before they reach customers.
Benefits of Moderators
- Catch AI mistakes before customers see them
- Maintain brand voice and quality standards
- Safely roll out AI Auto-Reply with controlled testing
- Edit responses before sending if needed
- Gather feedback to improve AI training
Adding Moderators
- Click Assign in the Moderators section
- Select an employee from the dropdown
- Click Add Moderator to confirm
- Repeat to add multiple moderators
Review Queue
- AI-generated responses go to a review queue
- Assigned moderators see pending responses
- They can Approve, Reject, or Edit
- Approved responses are immediately delivered
Managing Moderators
- Click Manage to view current moderators
- Remove moderators using the trash icon
- Add new moderators using the dropdown
- Changes take effect immediately
Context Management
Context is the knowledge you provide to help AI generate accurate, on-brand responses. It includes company information and real conversation examples.
Company Context
General information about your company that guides all AI responses.
- Company name & industry
- Brand tone & voice
- Company policies
- Service guidelines
Conversation Context
Sample conversations showing how your team responds to common questions.
- Billing question + answer
- Technical issue + solution
- Feature request + response
- Complaint + resolution
Tip: Start with 5-10 conversation examples and 1-2 company context items. As you gather more examples, the AI gets smarter. Regularly update context based on new conversation patterns.
Best Practices
Do
- Start with moderators enabled for safe rollout
- Provide comprehensive company context upfront
- Include diverse conversation examples
- Set narrow conditions initially, broaden over time
- Monitor moderator feedback and improve context
- Disable image replies if not needed (save tokens)
Don't
- Deploy without moderators on day one
- Leave context empty and expect good responses
- Reply to complex or sensitive issues with AI only
- Enable image replies without token budget
- Ignore moderator feedback and quality issues
- Forget to activate context items
Troubleshooting
AI Auto-Reply not responding to messages?
- Check if the main toggle is enabled
- If conditions exist, verify the message matches them
- Ensure you have company context set up in Knowledge Hub
- If moderators are assigned, check the review queue for pending responses
Responses are inaccurate or off-brand?
- Add more company context (tone, industry, policies)
- Add conversation examples showing correct responses
- Ensure context items are marked as Active
- Have moderators edit responses to improve training data
High token usage?
- Check if 'Reply to Images' is enabled (uses 60,000–100,000 tokens per image)
- Disable image replies if not needed
- Set more specific conditions to reduce reply frequency
- Monitor the token usage dashboard in settings
Moderators not seeing responses to review?
- Ensure they are assigned in the Moderators section
- Check if they have access to the AI Auto-Reply review dashboard
- Verify conditions match incoming messages
- Check the main toggle is enabled