Self Service & Automation · Topic 4 of 4
Virtual Agent
ServiceNow Virtual Agent: conversational topics, Virtual Agent Designer, NLU intents, channels, live agent handoff and where it fits for the CSA exam.
- 4 min
- Virtual Agent is a conversational interface, built from topics.
- Virtual Agent Designer is where topics are created.
- A topic needs to be published to be used.
What it is
Virtual Agent is the chatbot of the platform. Users type a question or request, and the bot answers, guides them through a conversation, or hands them over to a person.
Its job is to resolve common requests without an agent: finding an answer, resetting something, checking a request, or starting a catalog order. Administrators build the conversations; users see them in chat.
How it works
Topics
A topic is one conversation the bot can have, for example "Check my request status" or "Find a knowledge article". Topics are built in the Virtual Agent Designer, a visual canvas where you add nodes and connect them in order.
A topic must be published before the bot can use it. Editing a draft does not change what users see.
Nodes
| Node type | What it does |
|---|---|
| Bot response | The bot says something to the user: text, a picture, a link or a list |
| User input | The bot asks a question and stores the answer, such as a choice or a reference to a record |
| Utility | Behind the scenes work: a decision, a script, a record lookup or a call to an automation |
The flow of the conversation follows the nodes in order, branching on decisions based on what the user answered.
Finding the right topic
The bot decides which topic to start in one of two ways:
| Method | How it works |
|---|---|
| Keywords | The topic is chosen when the user message matches words set on the topic |
| NLU (Natural Language Understanding) | A trained model matches the meaning of the sentence to an intent, which is tied to a topic |
NLU models are created and trained in the NLU Workbench, where you list intents and sample sentences called utterances. Better and more varied utterances give better matching.
Channels and handoff
The same bot can serve several channels, for example the Service Portal or Employee Center chat, and messaging tools such as Slack or Microsoft Teams. Users meet the bot where they already work.
When a topic cannot resolve the problem, it can transfer the chat to a live agent. The conversation history goes with the handoff, so the person does not ask the same questions again.
Building and testing a topic
A typical topic starts with a trigger phrase or intent, asks the user for any detail it needs, looks something up or performs an action, and ends with a clear result. A good topic finishes with a closing step that asks whether the problem was solved.
Test the topic in the Designer before you publish it. Try several phrasings, including ones with typing mistakes, to see that the right topic starts and that each branch reaches an end.
Who can build topics
Building and publishing topics is an administrator task and is limited to administrators with the appropriate Virtual Agent role. Ordinary users only chat with the bot.
Connecting to other features
Virtual Agent is a front door to features you already configured. A topic can search Knowledge and show an article, check or create records, or use a flow to complete a task. That is why the bot is in the self service chapter.
How the exam asks about this
- "Users should be able to check their request status by chat." Build a topic in Virtual Agent Designer.
- "The bot picks the wrong topic for natural phrasing." Add more utterances to the NLU model for the intent.
- "A user is not satisfied with the bot." The topic should offer a live agent transfer.
What to remember for the exam
- Virtual Agent is a conversational interface, built from topics.
- Virtual Agent Designer is where topics are created.
- A topic needs to be published to be used.
- Nodes are the steps: bot response, user input and utility.
- NLU matches meaning to intents using trained utterances; keywords match words.
- Live agent handoff is the exit when automation cannot resolve the request.
Do not confuse the NLU Workbench with the Virtual Agent Designer. The Workbench trains intent matching; the Designer builds the conversation steps.
Common mistakes
Leaving a topic in draft. The topic is complete in the Designer, yet users never reach it because it is not published.
Training with few utterances. A model with a handful of similar sentences fails on how real users phrase requests.
Building a dead end. A conversation that cannot hand off leaves users stuck when the bot does not resolve their request.
Check yourself
Where do you build the steps of a conversation?
In the Virtual Agent Designer, as a topic made of connected nodes.
How does NLU decide which topic to run?
A trained model matches the user's sentence to an intent, and the intent is tied to a topic.
What lets a user reach a person when the bot cannot help?
A live agent handoff configured in the topic.
Topic 4 of 4 in Self Service & Automation
Test yourself on Self Service & Automation
The full ServiceNow CSA question bank has 600 questions, with timed practice exams and an explanation for every answer.
This study guide is independent and is not affiliated with, endorsed by, or sponsored by ServiceNow. Vendor names, certification names and exam codes are trademarks of their respective owners and are used for identification only.