Advantages of natural language models over keywords

  • Release version: Australia
  • Updated March 12, 2026
  • 1 minute to read

    • Natural language models help Virtual Agent to process human language based on context and your company's data. In this way, what the user needs can be more accurately matched with a corresponding topic. Virtual Agent supports large language models (LLMs) and Natural Language Understanding (NLU).

    Language is difficult

    Keyword matching has its limitations. For example, sometimes an apple is a piece of fruit, and sometimes it's an electronic device. Context matters, and so does intent. Natural language models are designed to deal with the following problems:
    • There are multiple ways to describe the same thing.

      Examples: Office password reset or Reset my password for Office

    • Expressions can be ambiguous.

      Example: Server reporting mail missing after migration. What is missing, the server or the mail?

    • Contextual information is essential.

      Example: activate London instance in staging

    • Words can acquire new meanings over time.

      For example, a cell can pertain to biology or a cell phone.

    • Slang, acronyms, and industry idioms can be difficult to interpret.

      Example: set up SSO on the dev instance

    • Error messages are often hard to understand.
    Virtual Agent provides two kinds of natural language topic discovery. You can use both in your instance, but only one at a time in any given chat.
    LLM topic discovery in Virtual Agent
    Use LLMs to discover topics and access generative AI capabilities without building complex models, intents, or entities.
    Natural Language Understanding (NLU) topic discovery in Virtual Agent
    Use ServiceNow NLU or a supported provider to discover topics.