When AI Escalates to Human
Understand why the AI escalates conversations and how to handle them effectively.
What Is Escalation?
Escalation happens when the AI determines it cannot handle a customer request confidently and passes the conversation to you.
In the Inbox, you'll see:
- 🚨 "Requires Human" flag on the conversation
- Highlighted in the "Attention" filter
- Handoff message sent to customer (configurable)
- AI stops responding automatically
Why AI Escalates
1. Low Confidence
When: AI's confidence score falls below your threshold
Examples:
- Ambiguous request: "I need something for my hair"
- Conflicting information: "I need a cut... or maybe color"
- Unclear timeframe: "Sometime soon"
- Vague service request: "Whatever you recommend"
Why it's good:
- Prevents AI from making mistakes
- Ensures customer gets accurate help
- Protects your reputation
What you can do:
- Lower confidence threshold if escalating too often
- Improve service descriptions to help AI understand
- Add common variations to Knowledge Base
2. Complex Requests
When: Request requires human judgment or expertise
Examples:
- "I have very specific color requirements based on my skin tone"
- "I need to discuss my medical conditions before booking"
- "Can you do this custom service I saw on Instagram?"
- "I want multiple people booked at the same time"
Why it's good:
- Prevents inappropriate bookings
- Ensures customer gets proper consultation
- Allows for custom solutions
What you can do:
- These escalations are appropriate
- Respond with consultative approach
- May require in-person consultation before booking
3. Policy or Complaint Questions
When: Customer asks about policies, terms, or has complaints
Examples:
- "What's your refund policy?"
- "I had a bad experience last time"
- "Do you accept insurance?"
- "I want to file a complaint"
- "Can I get a discount?"
Why it escalates:
- Requires business owner authority
- May need case-by-case judgment
- Sensitive customer service issue
What you can do:
- Add common policy questions to Knowledge Base
- Handle complaints with empathy and professionalism
- Document resolution for future similar cases
4. Out-of-Scope Requests
When: Customer requests something you don't offer
Examples:
- Service not in your list
- Outside your service area or location
- Requesting products you don't sell
- Asking for services on days you're closed
Why it escalates:
- AI can't book something that doesn't exist
- Might require alternative suggestion
- May need to explain limitations
What you can do:
- Politely explain what you do offer
- Suggest similar services if applicable
- Refer to competitors if appropriate
5. Technical Issues
When: System encounters a problem
Examples:
- No availability found (all slots booked)
- Service or location not active
- Integration issue
- Webhook or API error
Why it escalates:
- AI cannot complete the booking
- Prevents customer frustration with errors
- Allows you to handle manually
What you can do:
- Check calendar for actual availability
- Verify service/location settings
- Check integration status
- Contact support if persistent
6. Customer Specifically Requests Human
When: Customer explicitly asks to speak with a person
Examples:
- "I want to talk to a real person"
- "Can I speak to the owner?"
- "This bot isn't helping"
- "Connect me to someone"
Why it happens:
- Customer preference
- Frustrated with AI interaction
- Wants personal touch
- Trust or rapport concerns
What you can do:
- Respond promptly and personally
- Acknowledge their preference
- Build rapport through conversation
- Complete their request efficiently
7. Conversation Limit Reached
When: Message exchange exceeds configured limit
Default: 10 messages
Why it happens:
- Prevents endless back-and-forth
- Customer may be confused
- Request is more complex than AI can handle
What you can do:
- Review conversation to understand issue
- Take over and clarify directly
- Adjust conversation limit if appropriate (Settings)
Identifying Escalated Conversations
In the Inbox
Conversation List:
- 🚨 Red "Requires Human" flag
- Appears at top of list (priority)
- Visible in "All" and "Attention" filters
Using Filters:
- Click Attention filter
- Shows only conversations needing your action
- Clears when you respond
Unread Badge:
- Red number on Inbox navigation
- Indicates total unread (including escalations)
Email/SMS Notifications
If enabled (Settings → Notifications):
- Email alert: "AI has escalated a conversation"
- SMS alert: "Customer needs your attention"
- Includes customer name and message preview
Response time expectation:
- Configure in handoff message template
- Default: "Someone will get back to you shortly"
- Recommended: Respond within 1 hour during business hours
Handling Escalated Conversations
Step 1: Review Context
Before responding, check:
Read full conversation
- What did customer ask for?
- What did AI already say?
- Where did the conversation break down?
Check AI confidence
- Hover over messages to see confidence score
- Low scores indicate uncertainty
Understand escalation reason
- Scroll to see escalation point
- AI usually provides reason in system message
Step 2: Formulate Response
Good response structure:
Acknowledge - Show you're a human and you're here to help
- "Hi [Name], this is [Your Name] from [Business]"
- "I'm here to help you personally"
Clarify - Ask specific questions if needed
- "To make sure I book the right service, can you tell me..."
- "Are you looking for [Option A] or [Option B]?"
Provide solution - Give clear next steps
- "I have availability at [times]"
- "Let me book that for you right now"
- "I recommend we do a brief consultation first"
Complete action - Follow through
- Create appointment manually if agreed
- Send confirmation
- Set expectations for next steps
Example response:
Hi Sarah, this is Maria from Bella Salon! 👋
I see you're interested in a color service. To give you the best result,
I'd like to know a bit more about what you're envisioning.
Are you looking for:
• All-over color
• Highlights/lowlights
• Balayage or ombré
• Color correction
Also, have you colored your hair recently? This helps me recommend
the right service and timing.
Looking forward to helping you!
Step 3: Take Action
If booking confirmed:
- Go to Calendar
- Click "New Appointment"
- Fill in details from conversation
- Save appointment
- Notify customer of confirmation
If more info needed:
- Continue conversation until clear
- Ask clarifying questions
- Provide expert recommendations
- Book once agreed
If cannot fulfill:
- Explain why clearly and kindly
- Offer alternatives if possible
- Refer elsewhere if appropriate
- Thank them for understanding
Step 4: Close Conversation
After resolving:
- Confirm booking details with customer
- Provide any additional info (parking, what to bring, etc.)
- Invite further questions
- Mark conversation as read
Example closing:
You're all set! 🎉
Your appointment:
📅 Thursday, March 15 at 2:00 PM
💇 Balayage Hair Coloring (3 hours)
📍 123 Main Street
Please arrive with clean, dry hair.
See you Thursday! Let me know if you have any questions.
Preventing Unnecessary Escalations
Update Knowledge Base
When you see patterns:
- Same question escalated multiple times
- Common policies AI doesn't know
- Terms customers use that AI doesn't recognize
Action:
- Go to Settings → AI & Automation → Knowledge Base
- Add the information
- Test with a sample message
Example:
- Pattern: "Do you do beard trims?" escalates 3 times
- Your services: "Men's Haircut" (includes beard trim)
- Knowledge Base update: "Our Men's Haircut service includes beard trimming"
Adjust Confidence Threshold
If escalating too frequently:
- Review escalated conversations
- Identify if they were actually simple requests
- Lower threshold by 5% (Settings → AI & Automation)
- Monitor for 1 week
- Repeat if still escalating too much
Caution:
- Don't lower below 60% (too risky)
- Monitor for mistakes when lowering
- Raise again if quality suffers
Improve Service Descriptions
When AI consistently misunderstands services:
- Go to Services
- Edit service
- Enhance description with:
- What's included
- What's not included
- Who it's for
- Common alternative names
Enable AI Learning
Settings → AI & Automation:
- Turn ON "Enable AI Performance Learning"
- System will detect patterns automatically
- Apply synonym associations
- Improve over time without manual intervention
Providing Feedback on Escalations
When AI Shouldn't Have Escalated
If request was simple enough for AI:
- Handle the conversation
- Click "AI Feedback" button
- Select "Other"
- Describe: "This was a straightforward booking request. AI could have handled it."
- Submit
Impact:
- Helps AI learn when to be more confident
- Improves resolution rate over time
- Reduces your workload long-term
When AI Should Have Escalated Sooner
If AI struggled with something complex:
- Take over conversation
- Click "AI Feedback" button
- Describe what went wrong
- Submit
Impact:
- AI learns to escalate similar situations faster
- Prevents frustrating customer experiences
- Improves AI's judgment
Setting Response Time Expectations
Handoff Message Template
Location: Settings → AI & Automation → Human Handoff Message Template
Default: "I've flagged this conversation for our team. Someone will get back to you shortly!"
Recommended customization:
Let me connect you with my colleague [Your Name] who can help you better!
Our typical response time:
• Business hours (9 AM - 6 PM): Within 1 hour
• After hours: Next business day
For urgent matters, call us at {{phone}}.
Managing Customer Expectations
Be realistic:
- Don't promise instant response if you can't deliver
- Set specific timeframes
- Provide alternative contact for urgency
Overdeliver:
- If you say "1 hour", try for 30 minutes
- Customers appreciate faster-than-expected response
Common Escalation Scenarios
Scenario 1: Price Negotiation
Customer: "Can I get a discount?"
Why escalated: Pricing decisions require owner approval
How to handle:
- If you offer discounts: Explain conditions
- If you don't: Politely decline, emphasize value
- Consider first-time customer offers
Scenario 2: Special Accommodations
Customer: "I'm in a wheelchair, is your location accessible?"
Why escalated: Requires specific factual information
How to handle:
- Provide honest assessment
- Describe accessibility features
- Offer assistance if needed
- Add to Knowledge Base for future
Scenario 3: Group or Event Booking
Customer: "Can I book appointments for my wedding party (8 people)?"
Why escalated: Requires coordination beyond single booking
How to handle:
- Assess capacity and timing
- Propose schedule
- Discuss package pricing if applicable
- Create multiple appointments manually
Scenario 4: Service Modification
Customer: "Can you do X service but add Y from a different service?"
Why escalated: Custom service configuration
How to handle:
- Determine if feasible
- Calculate appropriate pricing
- Set clear expectations
- May need multi-service booking
Scenario 5: Complaint or Dissatisfaction
Customer: "My last appointment was terrible"
Why escalated: Requires empathy and resolution authority
How to handle:
- Apologize sincerely
- Ask for specific details
- Offer resolution (discount, redo, refund)
- Document for quality improvement
- Follow up afterward
Related Articles
- How AI Agent Works - Understanding AI fundamentals
- Training Your AI - Reduce unnecessary escalations
- AI Performance Monitoring - Track escalation patterns
Escalations are opportunities to provide exceptional personal service! 🤝