Getting started with Virtual Agent Designer
Summarize
Summary of Getting started with Virtual Agent Designer
The Virtual Agent Designer is a diagram-based tool for ServiceNow customers to create and manage conversation topics that guide users through self-service tasks or resolve common issues. Topics serve as blueprints for conversations between virtual agents and users, incorporating components like user inputs, bot responses, subflows, and actions. It supports both Natural Language Understanding (NLU) and Large Language Model (LLM) methods for matching user requests to the appropriate topics, enhancing user experience and efficiency.
Show less
Key Features
- Topic Creation and Management: Design topics with various controls, manage subflows and actions, and use the Assistant Designer Asset library to organize assets such as topics, topic blocks, small talk, custom controls, and AI agents.
- LLM and NLU Integration: Use generative AI (LLM) through Now Assist in Virtual Agent for topic discovery and matching, or leverage NLU keyword/topic discovery by mapping topics to intents within NLU models directly from Virtual Agent Designer.
- Assistant Selection and Configuration: Filter and assign assistants like Now Assist in Virtual Agent or create new assistants for conversational interfaces to optimize topic handling.
- Topic Properties and Flow Design: Configure topic attributes such as purpose, access controls, channels, and associate live agent variables. Build conversation logic visually via diagram or table views, enabling detailed flow control and script variable integration.
- Multilingual Support: Manage translation status, edit translations, and handle language mappings for topics to support multiple languages and improve accessibility.
- Testing and Debugging: Preview, test, and debug both LLM-based and NLU/keyword topics within the Virtual Agent Designer’s chat window to ensure correct behavior and intent matching.
- Asset Promotion and Ordering: Promote conversational subflows, actions, and topics to be suggested by LLM assistants and reorder them to optimize recommendation priority.
- Conversational Settings and Management: Manage individual asset settings, conversational subflows, actions, custom skills, and AI agents centrally through the Assistant Designer Asset library.
Practical Benefits for ServiceNow Customers
- Create tailored virtual agent conversations that improve user self-service and reduce workload on live agents.
- Leverage advanced AI technologies to enhance topic discovery accuracy and conversation relevance, boosting user satisfaction.
- Seamlessly manage and update conversational assets in one interface, ensuring consistent and efficient virtual agent maintenance.
- Support multilingual user bases with integrated translation features.
- Test and debug topics in real time to deliver reliable conversational experiences.
- Efficiently migrate existing NLU topics to modern LLM topics to take advantage of generative AI capabilities.
The Virtual Agent Designer is a diagram tool for creating and managing topics, which are blueprints for conversations between a virtual agent and user. You can design topics that help your users resolve common work issues or guide them through self-service tasks. You can also view and manage conversational subflows and actions for LLM-enabled conversations in the Assistant Designer Asset library.
Virtual Agent conversations are built using topics. When building a topic, there are various components (controls) that you can use to build the logic of a conversation, as well as the responses that the user sees. Virtual Agent conversations can also include subflows and actions if they were made conversational in Workflow Studio or topics that include subflows and actions through the Action utility node. Conversational subflows and actions are only available in Assistant Designer if Now Assist in Virtual Agent has been configured. Conversational subflows and actions only appear to users during a conversation if the Subflows and actions skill is turned on in Now Assist Admin. For more information, see Now Assist in Virtual Agent.
Virtual Agent can use either generative AI or Natural Language Understanding (NLU) to match the user's request with the appropriate conversation or topic.
With large language model (LLM) topic discovery, you can use generative AI through Now Assist in Virtual Agent to match the user's request with the desired conversation or topic. With Now Assist in Virtual Agent, you can test and publish your models from within Virtual Agent Designer. Now Assist in Virtual Agent uses LLMs and generative AI skills to improve deflection rates and reduce the amount of time-consuming work that Natural Language Understanding (NLU) topic discovery requires. For more information, see LLM topic discovery in Virtual Agent.
With NLU/keyword topic discovery, the topic author associates the topic with an intent within an NLU model. An intent describes what the user wants to achieve. User utterances are matched with intents to determine the most appropriate topic for the user. Associating a topic with a model and intent is called mapping.
If you're using ServiceNow NLU/keyword topic discovery, then Virtual Agent Designer integrates with NLU Workbench so that you can map topics to new or existing models and intents without leaving the interface. In addition, you can also modify utterances, test, train, and publish your models from within Virtual Agent Designer.
Assistant Designer Asset library
You can test and preview active topics to verify that they work as intended. You can also sort and search for topics. When you have many assets, use
the search bar and filter icon
to quickly organize and find your topics. For more information, see Assistant Designer Asset library.
| Field | Description |
|---|---|
| Name | Name of the topic. Template topics also contain (Template) in the name. |
| Type | Topic types include Topic, Topic Block, Setup Topic, Small Talk, Custom Input Control, Custom Response Control, Subflows, Actions, and AI agents. |
| Status | Status types include Draft, Modified (with unsaved changes), or Published. |
| Active | Whether the asset is active or inactive. |
| Last modified | Time (minutes, hours, days, or months) that the asset was last modified. |
| Description | Description of the asset. For topics, this field is created and updated on the Properties tab. |
When you open a topic, a tab is displayed in the navigation header bar. The topic's name is shown on the tab, and a red dot appears if the topic has unsaved changes. Hover over the tab to view the scope in which the topic was created.
When you open a subflow or action, a tab is displayed in the navigation header bar. This tab opens the subflow or action in Workflow Studio within the Virtual Agent Designer environment.
Assistants
- Now Assist in Virtual Agent (default)
- Now Assist Panel - Platform (default)
You can create new assistants from . For more information about creating assistants, see Create a chat assistant.
For more information about LLM assistants, see LLM assistants.
Topic properties
Use the topic Properties tab to identify a Virtual Agent asset and how it's used. The properties you can specify depend on the method of topic discovery used in your instance: LLM or NLU/keyword. For example, if you have activated Now Assist and are working with an LLM topic, the Properties tab includes fields for what you want the topic to do and which assistants to use. If you have activated NLU, you can select NLU/keyword from the Type drop-down menu when you create a topic. The Properties tab includes fields for identifying the NLU model and intent for the topic.
- Specify the name and type of asset, such as topic, topic block, custom control, and so on.
- Specify the model type, purpose of the topic, and which assistants can run this topic for Now Assist customers.
- Specify the model type, add a topic description, specify the purpose of the topic, and select which assistants can run this topic for Now Assist customers.
- Control who uses the topic and what channels it can run in.
- Associate live agent variables with a topic.
- Add keywords to the topic or map the topic to a ServiceNow NLU model and intent.
For more information, see Topic Properties tab.
Topic flow
When you create or update a topic, topic block, or custom control, you build the conversation flow on the Flow tab. In the default view, the flow looks like a diagram. Use the controls for user inputs, bot responses, and utilities to define the flow. The status of the topic and NLU model (if used) appears in the upper-right corner of the window.
The Table View option in Virtual Agent Designer displays the nodes in your conversation flow as a table, rather than as a diagram. Each node is identified in a corresponding row in the table. You can add, change, or delete nodes using these table rows.
For example, instead of dragging controls and dropping them onto the canvas, you can select Add new node to insert a node into the appropriate row in the table. When you focus on a particular row (node), the corresponding property sheet for the node is displayed.
As you create or update your design, you might find it helpful to switch between the diagram view and the Table View using the Table View option. If your conversation contains many nodes, the Table View lets you review all the nodes in your conversation quickly. You can use the Search table option to quickly find a particular node or control.
- Design your topic flow. For more information, see Design the flow of your topic.
- Add script variables. For more information, see Define script variables for a topic.
- Add entities to your topic. For more information, see Use system-derived entities in your topic.
- Add nodeless entities, if you’re creating a topic for Add nodeless NLU entities to your topic flow.
For more information, see Topic Flow tab.
NLU model and intent mappings
- Review utterances used in your topic. For more information, see Review and edit utterances and associated entities.
- Train your NLU models. For more information, see Train, test, and publish your NLU model.
Test the topic model with user utterances to see if it triggers the appropriate intent and topic. You can also test entity recognition.
For more information about this tab, see Topic NLU Intent tab.
Topic languages
The Languages tab displays the topic's translation status for languages that have been activated on the instance for both LLM and NLU conversations. If you're using NLU/keyword, you can also review the language mappings for that topic.
- View translation status for activated language plugins.
- Test your topic with a given language. For more information, see View and test the topic in a specific language.
- Edit translations. For more information, see Edit the topic translations.
- Request translations. For more information, see Request topic translations in additional languages.
- View and update NLU language and entity mapping. For more information, see View or modify NLU language mappings and entity mapping.
For more information, see NLU model mapping in Virtual Agent Designer.
For more information about this tab, see Topic Languages tab.