EN
Practical guide · Customer support and human relay

ChatBot: transfer a conversation to a human agent

Prepare for transfer to reception by the agent. Prepare for success and failure outputs, then confirm the actual pass in the published LiveChat widget.

Analysis of official documentation, with no testing on a paid account.

A conversation stream passes from a chatbot to a support agent through a transfer door.
Original illustration of a controlled relay between bot and human agent. This visual is not a capture of ChatBot or LiveChat.
In this guide
  1. Choose an understandable trigger
  2. Building the path of success and failure
  3. Check LiveChat side prerequisites
  4. Validate in two levels, not with a single green light
  5. Take this case: an order already paid
  6. When this assembly is not suitable

To transfer a ChatBot conversation to a human, plan two paths from the "Transfer chat" block: success to the agent, failure to a response that explains what to do next. The actual handover requires an active LiveChat integration, automatic assignment and at least one available agent accepting chats. Check the trigger in the internal testing tool, then confirm that an agent receives the conversation.

This guide aims at a small team that automates common issues without wanting to lock a client in a loop. Official aid ChatBot Consulted the 21 October 2026.

Choose an understandable trigger

Do not reserve the transfer to the technical failure of the bot alone. Also add an explicit way, such as a "Talk to a person" button, when the request relates to a decision that the bot must not make. A change of order already paid, a claim or a situation specific to the customer file often justify this exit.

Three triggers and their effect
TriggerActionWhy?
The visitor asks for a humanAdvertise the transfer and trigger the block.The exit remains visible and voluntary.
Two answers not foundUse the fallback chess counter before transferring.The visitor gets a second chance without a loop.
File specific requestTransfer before requesting unnecessary sensitive data.The agent takes the context and applies the internal procedure.

ChatBot’s help specifically suggests a filter after the fallback: the first failure displays a clarification response, then a count greater than or equal to two can announce a handover. This value is a documented starting point, rather than a universal rule. An ambiguous question may need rephrasing; a sensitive enquiry may require a human immediately.

Building the path of success and failure

Place the “Transfer chat” block at the chosen point and connect both outputs. The Success output ends the bot’s workflow: the human agent takes over the conversation, and responses placed after that output will not appear. The Failure output is equally important, as it activates in particular when no agent is available, agents are not accepting chats or their simultaneous chat capacity has been reached.

  1. Write a short message before the transfer: "I connect you with the team. »
  2. Choose the LiveChat group to receive the conversation.
  3. Connect Success at the end of the course, without message from the bot supposed to appear after recovery.
  4. Connect Failure to a response that clearly says no one can take the chat right now.
  5. Then propose a real way to leave the request: ticket configured, form or address followed by the team.

Do not promise a response time you do not measure. The failure path can say “The team is unavailable to take over this chat. Leave your enquiry here” and link to the channel you handle. Organise an asynchronous channel for enquiries that arrive when agents are absent.

Check LiveChat side prerequisites

The transfer action is designed for LiveChat integration. The vendor indicates that the automatic assignment must be enabled; manual selection prevents the transfer described in this path. Also check the chosen group and its routing rules, as a LiveChat rule can run the conversation somewhere else.

Before publishing the bot, record for each group who may accept chats, at what times, with what simultaneous capacity and which channel takes over when nobody is available. This internal record prevents confusing a well-placed block with an organisation that is ready to respond.

Validate in two levels, not with a single green light

What each check shows
CheckWhat it validatesWhat it doesn't validate
Test tool ChatBot in debug modeThe message reaches the right block and triggers the intended action.The chat reaching a LiveChat agent.
Widget LiveChat published, agent availableThe transfer passes through Success and the agent receives the conversation.Behaviour when the team is unavailable.
Widget LiveChat, agent not available or capacity reachedThe Failure path displays the expected message and relay.The speed of future pick-up of the ticket or form.
LiveChat archivesFull conversation before and after transfer.The quality of the agent's response.

The manufacturer's modifications become active only after the new version of the bot is published. So do the internal test for routing, publish, and then run two separate conversations in the installed widget: one with an agent available and one without capacity available. In ChatBot, the archives stop at the passage to the agent; the complete history is consulted side LiveChat.

Take this case: an order already paid

A customer wrote: "I want to change the address of an order already paid. The bot acknowledges that this action depends on a folder and displays: "This change must be verified by the team. I'll put you in touch. It then triggers the transfer to the command group.

If an agent is accepting chats, the Success output ends the automated workflow and the agent takes over with the context. If no agent can take the conversation, the Failure output displays: “No one can take over this chat right now. Send your order number and new address through our secure form.” This example assumes that the form exists and someone monitors its queue; otherwise, replace it with the available channel.

Control is successful only if the first case appears in LiveChat with the right group and if the second one follows the path of failure without promising an immediate recovery. A "action triggered" message in the test tool is not this proof.

When this assembly is not suitable

This route is not suitable if you do not use LiveChat, if no one accepts conversations or if you need an asynchronous messaging managed solely by the bot. Documentation also indicates that the bot does not work with the asynchronous messaging mode in this transfer scenario.

For an emergency, a security incident or a regulated request, directly display the appropriate channel and schedules instead of crossing a bot. For a team that deals mainly with deferred requests, a ticket with owner and take-back line can be more honest than a transfer presented as immediate.

To complete the course, see ChatBot review, the guide to receive messages off schedule and the method to assign each request HelpDesk.

View ChatBot →

Sources and method of this analysis

ChatBot: Set up transfer to an agent LiveChat ; ChatBot: test limits and record reading LiveChat ; ChatBot: test tool, simulation and publication of the bot ; ChatBot: user input and fallback trigger ; ChatBot: archive content and transfer trace Integration, groups, availability and routing must be verified on the accounts concerned. Sources consulted on 21 October 2026. Methodology · Trade transparency · Report correction.

Logicelio newsletter

The right tools, without spending hours comparing.

Receive new Logicelio guides and comparisons, along with offers and benefits from our partners when they are relevant to freelancers and small businesses.

  • New guides and comparisons
  • Partner offers and benefits
  • Free 2026 invoicing guide after signing up

Your email address is used for the newsletter. The 2026 invoicing guide is included as a free bonus after signing up. Privacy.