This form is used to set the properties for a new conversational topic for Virtual Agent.

The form used to start a new topic is nearly identical in layout to the Topic Properties tab.

Note: LLM topics are faster and easier to set up and configure compared with NLU topics. The LLM model type doesn't need you to select, configure, or train models or intents. For more information, see LLM topic discovery in Virtual Agent.

Two versions of the Create a topic form. First, the Model Type is NLU/Keywords. Second, the Model Type is LLM, where a description is required for topic discovery.

Field Description
Type

List of objects that you can create. Select Topic.
Name Unique name for the topic that reflects its business purpose.
Model Type

Select LLM to use topic discovery with LLMs, or select NLU/Keywords to use either keyword or NLU discovery. For more information, see LLM topic discovery in Virtual Agent.

Note: If this field doesn't appear, ensure that Now Assist in Virtual Agent is turned on. For more information, see Configuring Now Assist in Virtual Agent.
LLM topics

[LLM topic discovery only]
What do you want the topic to do? A detailed explanation of the LLM topic's purpose to differentiate it from other topics. For example: Help users see how many vacation days they have left.
Assistant(s) The Now Assist in Virtual Agent assistant that the LLM topic or small talk topic is associated with for more accurate topic discovery. For more information, see LLM assistants.
Natural Language Understanding (NLU) topics

[NLU/Keywords topic discovery only]
Description Brief explanation of the topic purpose and functionality.
If you use single quotes (‘ ’) in your description, do the following to avoid problems with your conversation:
  1. Navigate to All, then enter sys_properties.list in the filter.
  2. Search for and select glide.ui.escape_all_script, then set its value to True.
NLU Model Natural Language Understanding model to be mapped to the topic. The model contains the user utterances that are associated with the trained intents and entities. This combination enables your bot to understand what a user says. Do one of the following:
  • Choose an existing model.
  • Select Create Model to create a new model group.
    1. In the Model Name field, enter a name for the new model group.
    2. In the Intent name field, accept the default value or enter a name for the new intent.
      Note: New models are created in Draft state.
Tip: To see how and where the NLU model is used, navigate to NLU Workbench > Models, and sort on the Created for column.
Associated Intent The intent, when recognized by Virtual Agent, that launches this topic. If you specified or created a model, the intent is required. Do one of the following:
  • Choose an existing intent.
  • Select Create intent to create a new intent.

    In the Intent Name field, accept the default value or enter a name for the new intent.

    Note: This option is not available for read-only models.
Enable DialogActs Toggle switch (active by default in new topics) to enable Virtual Agent to detect when a user modifies utterances or applies updates in an NLU model, and dynamically reacts to unexpected utterances and modifications. Slide this toggle to the off position to prevent users from using the keyboard instead of the options available in the topic. For more information, see Dialog Acts for Virtual Agent.
Note: This toggle is decoupled from the topic switching option found in all User Inputs aside from Grouped Choice and File Picker. It is available for all topic types aside from a Topic Block. This feature is currently available only for English.
Confirm modified values with users Toggle switch to enable Virtual Agent to send a confirmation message to the user when the Virtual Agent detects a value change. This value is active by default.
Note: This toggle appears when Enable DialogActs is activated.
Advanced properties (optional)
Who can access this topic
System Roles

Roles that a user must have to view and run the topic. If a topic is public (available to users, including guest users, who aren't authenticated in the ServiceNow platform), select only the Public role.
Channels

Chat clients in which the topic or topic block will run.

These channels are different from the default channels that are natively supported by Virtual Agent. Your instance may have additional channels for custom chat integrations created for it. For more information about channels, see Deploying Virtual Agent topics in other channels.

Applicable to only keyword or NLU topic discovery.
Context
Expression logic to specify a condition for presenting this topic to users. The condition must evaluate to true. You can set this field using either the condition builder or a script.
  • Condition: Use the condition builder to add or edit conditions.
  • Script: Use the script editor to create or edit a script that contains a condition statement.
Live agent
Live agent context variables Live agent context variables that are defined in the General Chat Settings.

Select the variables that provide dynamic context, which means information from the Virtual Agent chat. This information is transferred from the bot conversation to the live agent.

For information about the default list of live agent variables, see Live agent chat context variables.
Available for Agent Autopilot Toggle switch that determines whether the topic is available to a live agent.

When enabled, an agent can search for and invoke the topic. For details, see Conversation Autopilot.

Applicable to only keyword or NLU topic discovery.
Additional
Resume flow after topic switching

Toggle switch that lets the user return to the original conversation after changing topics during a conversation.

Applicable to only keyword or NLU topic discovery. Not applicable to Small Talk topics.
Categories Label used to identify and group-related topics, topic blocks, or custom controls.
Keywords

[NLU/Keywords topic discovery only]

List of key phrases or terms that users enter to initiate the conversation with the Virtual Agent. Press Enter after each phrase. Keywords are also used for languages that are currently not available in NLU.