Virtual Agent
TODO:FIXME
Jace Benson
Table of content
Virtual Agent is ServiceNow's conversational AI platform. It lets you build chatbots that handle service requests through natural language conversations — a user types "I need a new laptop" and the bot walks them through catalog selection, approval, and fulfillment without ever opening a form.
Virtual Agent runs on topics (conversation flows built in Flow Designer), NLU (natural language understanding to interpret what users mean), and conversation channels (web, mobile, Slack, Teams, etc.).
The platform now also supports Now Assist for Virtual Agent, which brings generative AI to conversations — summarization, suggested replies, and LLM-driven topic discovery.
Key Concepts
- Topics — The conversation flows that define how a bot handles a specific request. Built in Flow Designer, topics have a trigger, conversation steps, and resolution.
- NLU Models — Natural language models that map user utterances (what they type) to topics and entities (what they're talking about). ServiceNow provides pre-built models and you can train your own.
- Channels — Where the conversation happens. Web chat, mobile app, Microsoft Teams, Slack, WhatsApp, and custom channels via the conversation API.
- Now Assist — Generative AI integration for Virtual Agent. Automated topic creation from knowledge articles, conversation summarization, and LLM-powered responses.
Topics vs. Traditional Automation
| Aspect | Virtual Agent | Traditional |
|---|---|---|
| Trigger | User intent (NLU) | Form submission, record action |
| Flow | Conversation branches | Linear process |
| State | Session-scoped | Transaction-scoped |
| Output | Answer or ticket | Database record |