Human Handoff

Seamlessly take over conversations from AI when human expertise is needed.

Human handoff is the process of smoothly transitioning a conversation from AI to a human agent. It's the "human by design" part of AI Hybrid Chat.

When Handoff Happens

Handoff can occur in three ways:

1. AI-Initiated

The AI recognizes it needs help:

  • Confidence score below threshold
  • Complex question outside knowledge base
  • Customer expresses frustration
  • Topic requires human judgment

2. Customer-Requested

Customer explicitly asks for human help:

  • "I want to talk to a person"
  • "Can I speak with a human?"
  • Clicks "Talk to a person" button

3. Agent-Initiated

Your team proactively takes over:

  • Sees high-value customer
  • Notices sensitive topic
  • Wants to provide personalized help

The best handoffs are seamless — customers shouldn't feel like they're being "transferred" but rather that help is arriving.

How Handoff Works

Trigger occurs

One of the three handoff scenarios above happens.

Team is notified

Agents receive notification (Slack, email, push).

Conversation is queued

Appears in open/unassigned conversations.

Agent claims or is assigned

Team member takes ownership.

Human responds

Agent continues the conversation.

AI steps back

AI Copilot assists but doesn't auto-respond.

Taking Over a Conversation

From the Dashboard

Go to Conversations

See unassigned/open conversations.

Click 'Take Over'

Claim the conversation.

Review context

Read the conversation and AI summary.

Respond

Send your message to the customer.

From Notifications

If you receive a Slack or push notification:

  1. Click the notification link
  2. Opens directly to the conversation
  3. Review and respond

What Customers See

During handoff:

  1. Message: "Let me connect you with a team member..."
  2. Brief wait (if agents are available)
  3. Agent introduction: "Hi! I'm [Name], happy to help!"
  4. Conversation continues naturally

Customize the handoff message in your widget theme settings.

AI Copilot

When you take over, AI Copilot helps:

Response Suggestions

  • AI drafts suggested responses
  • Based on conversation context
  • Uses your knowledge base
  • You can accept, edit, or ignore

Context Summary

  • AI provides conversation summary
  • Key points highlighted
  • Customer intent identified
  • Related knowledge base articles shown

Knowledge Search

  • Search your knowledge base
  • Insert answers into responses
  • Link to documentation

Availability and Routing

Agent Availability

Set your status:

  • Available — Ready for handoffs
  • Busy — Currently handling chats
  • Away — Temporarily unavailable
  • Offline — Not working

Routing Rules

Configure how handoffs are routed:

MethodDescription
Round RobinDistribute evenly
First AvailableFastest agent takes it
Skill-BasedMatch by expertise
Load BalancedBased on current workload

Offline Handling

When no agents are available:

  • AI continues to help
  • Collects customer email
  • Queues for follow-up
  • Sends email when agent responds

Escalation Rules

Use orchestrations to control handoff:

Always Escalate

For sensitive topics:

Condition: message_contains: refund, cancel, legal
Action: escalate_to_human

Route to Specific Team

For specialized questions:

Condition: url_contains: /enterprise
Action: notify_team: enterprise-sales

Time-Based Routing

Different handling by time:

Condition: business_hours: false
Action: ask_for_email, queue_for_later

Handoff Best Practices

Quick Response

  • Respond within 2-3 minutes of takeover
  • Acknowledge the customer immediately
  • If you need time, say so

Read the Context

  • Review the full conversation
  • Note what AI already tried
  • Understand the customer's frustration level

Don't Repeat

  • Build on what AI said
  • Don't ask questions AI already asked
  • Reference previous answers

Be Human

  • Introduce yourself
  • Show personality
  • Express empathy when appropriate

Follow Through

  • Solve the issue completely
  • Confirm resolution
  • Offer additional help

Measuring Handoff Performance

Track in Analytics:

MetricTarget
Handoff Rate< 30% of conversations
Response Time< 3 minutes
Resolution Rate> 90% after handoff
CSAT after Handoff> 4.5/5

Common Issues

Too Many Handoffs

  • Improve knowledge base
  • Raise AI confidence threshold
  • Add orchestrations for common questions

Slow Response to Handoffs

  • Enable push notifications
  • Set up Slack alerts
  • Ensure adequate staffing

Awkward Transitions

  • Customize handoff messages
  • Train AI to summarize for agents
  • Ensure agents read context

Next Steps

Was this helpful?