Dante AI publishes this guide. Dante includes human handover and appears below as one option, so we have kept the framing criteria-first: judge any tool, ours included, against the same list.

Most guides to choosing an AI chatbot focus on the AI: the models, the training, the answers. This one focuses on the seam. An AI chatbot with human handover answers what it can and passes the rest to a person, and the quality of that pass is where tools separate. A customer escalated gracefully often leaves happier than one who never needed help. One dropped at the seam rarely comes back.

This is a buying guide, not an industry story. If you want to see handover working in a specific setting, we have written those separately: how handover lifts hotel guest satisfaction and how it is changing retail customer support. Here, the job is the criteria for evaluating any tool.

What Human Handover Actually Means

Human handover is the moment a live conversation moves from an AI agent to a person. Done well, it is invisible plumbing: the AI agent recognizes its limit or the customer asks for a person, your team gets notified, someone opens the conversation with the full transcript, and the customer keeps typing in the same window. Nobody explains their problem twice.

Done badly, handover is a trapdoor. The chat ends with an email address, a contact form, or a promise that someone will be in touch. The customer starts over, and everything the AI agent learned in the first five minutes is thrown away.

Why AI-Only Support Fails on the Edge Cases

A well-trained AI agent handles the bulk of routine questions: pricing, policies, how a feature works, order status. That is real value. But every business has conversations that need judgment. A complaint from an angry long-term customer. A refund request that sits outside policy but deserves a yes. A question with legal, medical, or safety weight.

An AI-only setup meets these moments in one of two bad ways: it guesses, which erodes trust, or it stonewalls, which loses the customer. The fix is not a smarter model but a clean route to a person, triggered at the right moment, with context intact. That makes handover a hard requirement, not a nice-to-have. Our guide on when to bring in a real person covers the judgment side in more depth.

The 7 Criteria That Separate Good Handover From a Dead End

Feature pages all say "human handover." These seven questions tell you what the phrase means in each tool.

1. Trigger rules

What causes the handover, and who decides? At minimum you want three triggers: the customer explicitly asks for a person, the AI agent fails to find a grounded answer, and the conversation touches a topic you have flagged as human-only. The tell is configurability. If you cannot adjust when escalation fires, you will escalate too often and drown your team, or too rarely and trap customers in a loop.

2. Context transfer

What does the person receive when they take over? The answer should be everything: the full transcript, any details the AI agent collected, and the page the visitor was on. If your team's first message after a handover is "can you tell me what you need?", the tool has failed this criterion, whatever the feature page says.

3. Notification channels

A handover nobody sees is a dead end with extra steps. Check how your team finds out a conversation is waiting: an inbox in the platform, email, a message where your team already works, or your existing help desk. Then check the speed. A notification that lands ten minutes late is a missed conversation.

4. Availability windows

Your team is not online at 3am, and a good tool is honest about that. Look for the ability to set the hours when handover to a live person is offered, and for sensible behavior outside them. The worst pattern says a person is coming when nobody is there. The best tells the customer the truth and moves to capture instead.

5. Transcript continuity

Does the conversation survive the handover as one thread? The customer should stay in the same window, and the record should show the whole journey, AI turns and human turns together. Continuity matters later too: reviewing transcripts should show the full story in one place, not stitched across two systems.

6. Fallback capture

When no one is available, the conversation should convert into a lead, not evaporate. The AI agent should collect the visitor's name, contact details, and question, set an honest expectation for follow-up, and put that capture somewhere your team will actually process it. This often decides whether after-hours traffic is an asset or a write-off.

7. Reporting

You cannot tune what you cannot see. The tool should show how often handovers happen, what triggered them, and which questions the AI agent keeps failing on. Rising escalations on one topic usually means a gap in your training content, cheap to fix once visible. If handovers are invisible in reporting, the product treats escalation as an afterthought.

How Dante AI Handles Handover

Stated plainly, so you can compare it like anything else. Dante trains an AI agent on your website and uploaded documents, and you embed it with a single script tag. Human handover is built in: when a conversation needs a person, your team can step in and take over with the full conversation in front of them, so the customer never repeats themselves. When nobody is free, lead capture collects the visitor's details and question so the conversation becomes a follow-up, not a dead end.

The free plan needs no credit card, which matters because handover is the kind of feature to test live rather than take on faith. Paid plans start at $40 per month in USD; details are on the pricing page. Dante is one good option, not the only one, and the checklist below works on any tool.

The Evaluation Checklist

Run this against every tool on your shortlist, ideally on a free tier with your own content loaded:

A tool that passes all seven is rare. One that fails context transfer or fallback capture should leave your shortlist, whatever else it does well. For selection questions beyond handover, see our guide to choosing AI customer service software and our small business chatbot comparison.

Frequently Asked Questions

What is human handover in an AI chatbot? The moment an AI chatbot passes a live conversation to a person, ideally with the full transcript attached and the customer staying in the same window.

Can an AI chatbot transfer a conversation to a live person? Yes, if the platform supports it. Favor tools where handover is built in rather than bolted on, and test the pass on a free plan before you commit.

When should an AI chatbot hand over to a human? When the customer asks for a person, when the conversation needs judgment (complaints, refunds, legal or safety topics), and when the AI agent cannot find a grounded answer.

What happens if no human is available when the chatbot escalates? A good tool falls back to lead capture: name, contact details, and the question, with an honest note about when to expect a reply.

Does human handover require a full help desk platform? No. Standalone AI chatbot platforms can include handover without seat-based help desk software. If you already run a help desk, check the chatbot routes into it.

The fastest way to know whether a tool's handover is real is to trigger one yourself. Start free with Dante, load your own content, ask for a person, and watch what happens at the seam. That one test will tell you more than any feature page.