Why use an Auto Escalation dialog?

The purpose of the Auto Escalation dialog is to free the consumer from being stuck within a question. This can happen when the bot doesn't recognize the consumer's input. Because the bot can't determine the next step that should happen, the fallback response is sent, and the consumer is returned to the place where the failure occurred. The fallback response is repeated after each unrecognizable input, resulting in a stuck conversation. The consumer's predicament is illustrated in the example below.

An example of a conversation where the consumer is stuck

The Auto Escalation dialog solves this problem by offering the consumer the option of being transferred to a live agent (or another bot). You configure the number of times that the fallback response should be sent within a question. Once that threshold is reached, the Auto Escalation dialog is triggered automatically.

An example of a conversation where the consumer is stuck but offered a way out, by being transferred to an agent


If you’ve implemented a KnowledgeAI integration in the Fallback dialog within the bot, don’t use an Auto Escalation dialog, as it will never be triggered by the system. Here’s why:

The purpose of the Auto Escalation dialog is to handle cases where the consumer repeatedly receives a fallback response like, “Sorry, I don’t understand” — in a repetitive loop — because the bot doesn’t understand the consumer’s response.

If you have a KnowledgeAI integration within the Fallback dialog, you’ve essentially solved that “bot problem” using a different solution, i.e., with a search of the knowledge base for an answer. So, an Auto Escalation dialog is never triggered if the Fallback dialog includes a KnowledgeAI integration of some kind.

If you require a way to escalate to an agent in cases where the knowledge base search repeatedly fails, you can accomplish this: In the KnowledgeAI interaction’s post-process code, define a botContext variable that keeps count of the number of loops. After a certain number of loops, use setTriggerNextMessage to trigger an Agent Transfer interaction.


Implement an Auto Escalation dialog

Step 1: Create the dialog

A bot can have only one Auto Escalation dialog.

To create an Auto Escalation dialog

  1. Open the bot, and click Add Dialog in the lower-left corner.
  2. In the dialog that appears, specify the following:
    • Dialog Name: Enter a descriptive name.
    • Dialog Type: Select "Auto Escalation Dialog."
    • Auto Escalation Skill: Specify the ID of the skill to which to escalate (transfer) the conversation. You can select from the IDs of the skills configured for your account; alternatively, enter a botContext variable like {$botContext.skillId} or an environment variable. You can edit this information later in the Escalation Integration interaction's settings.
    • Auto Escalation Threshold: Select the maximum number of consecutive times the fallback message should be sent within a question before triggering the Auto Escalation dialog. Example: You set the threshold to 3. The consumer enters an incorrect answer three times, receiving the fallback response each time. After the fourth incorrect answer, the Auto Escalation dialog is triggered.

      You can edit this information later in the dialog's settings. LivePerson recommends setting this to "3," but you can set this to a different value based on your confidence that the fallback response will resolve the user's issue.

  3. Click Save.

Step 2: Build out the dialog

By default, an Auto Escalation dialog includes a Yes/No multiple choice question that asks the consumer if they want to speak to a live agent. If the consumer says Yes, the escalation (transfer) occurs. If the consumer says No, the interaction ends.

The default configuration of the Auto Escalation dialog

If the default dialog implementation doesn't meet your requirements, build out and/or modify the dialog. For info on configuring Agent Transfer interactions, see this section.

When you create the dialog, you configure two important settings that you might want to change afterward:

Best practices

See these best practices for using Agent Transfer interactions. They include info on how to handle transfer failures and more.


Can you explain how the Auto Escalation Threshold counter works?

In general, the counter is incremented each time a failure response is sent for a question interaction. More specifically, this happens if any of the following occurs:

  • The built-in fallback response is triggered. This can be the one that's specific to the question or the built-in, default fallback response.
  • A fallback dialog that contains only text interactions exists and is triggered.
  • A fallback dialog that contains an API integration (e.g., a Knowledge Base integration) exists and is triggered. The integration call fails due to either zero results being returned or some other failure. The failure condition isn't handled via an API failure match condition, which triggers a system error message. It's the system error message that results in incrementing the counter.