Human Handover

Sometimes you need to pass control of the chat to a human operator to talk to your user. The Human Handover expected response immediately hands control of the chat to the messenger service you have integrated (such as Meta Messenger) for your human operator to control.

You can configure the handover to:

• Send a passage instead of handing the user over if it's outside your organisation's business hours.
• Provide a transcript summary to the human agent. You can configure:
  • A predefined summary using variables and constants.
  • A generated summary using an LLM Usage.

As the handover does not require or wait for your user to respond, you should use the Bot Message to inform the user they are about to be transferred to a human. If you want to ask the user whether they want to talk to a human, you should create a separate passage for that, where the "yes" option flows to a passage with the Human Handover expected response.

Send a passage when outside business hours

By default, the handover option immediately hands the user over to a human agent. You can configure the handover to send a passage instead when it's outside your organisation's business hours.

1. Make sure you have configured business hours for your organisation.
   The business hours should be relative to your chatbot's timezone. See General settings.
2. Select the Human Handover response.
3. Select Handover only during Business hours.
4. Select the set of Business hours to use.
5. Select the Passage or logic to send the user when it's outside business hours.
6. Click Save.

Provide a transcript summary

You can provide a summary of the chat transcript when handing a user over to a human agent. The summary helps the agent quickly understand the user's situation without having to read the whole conversation.

By default, the human handover option does not provide a transcript. You can configure:

• A pre-defined summary using variables and constants.
• A generated summary using an LLM Usage.

Add a predefined summary

You can create a predefined summary to provide the human agent. Predefined summaries can use variables, constants, and any entities that were used in the passage's training phrases, to add dynamic information.

You can test your summaries using your Testbot logs.

To add a predefined summary:

1. Select the Human Handover response.
2. Scroll down to the transcript summary section.
3. Select Predefined transcript summary.
4. Type the summary to send to the agents.
   Type @ to add variables, constants, or entities. Only entities that were used in the passage's training phrases are available.
5. Click Save.

Add a generated summary

A generated summary uses an LLM Usage to generate a custom summary unique to each user's interaction with your chatbot.

The summaries use the prompt defined in your primary LLM Usage for handover transcript summaries. By default, this prompt generates a summary in the format of: Purpose, Agent, Actions, Outcome, and Sentiment, which indicates whether the user was happy with the bot's responses.

You can test your generated summaries using your Testbot logs. Generated summaries may take several seconds to appear.

To add a generated summary:

1. Configure a primary LLM Usage for Handover transcript summaries:
   1. Create an LLM Connector
   2. Create an LLM Usage for the Handover transcript summaries scope.
   3. Set the new LLM Usage as Primary for Handover transcript summaries.
2. Open the passage where you want to add a generated summary.
3. Select the Human Handover response.
4. Scroll down to the transcript summary section.
5. Select Custom LLM-generated transcript summary.
6. If you want to customise the default prompt of the LLM usage for this passage, type any additional instructions.
   These additional instructions only apply for handovers initiated from this passage. To create customised instructions that apply to all handovers from your chatbot, edit the default prompt of your primary LLM Usage. Click the Primary prompt link to see the default prompt in a new tab.
   Type @ to add variables, constants, or entities. Only entities that were used in the passage's training phrases are available.
7. Click Save.