Questions and responses in an exploration

  • Release version: Australia
  • Updated March 12, 2026
  • 4 minutes to read

    • Ask the AI specific questions in AI Data Explorer, to which it responds with data visualizations, a summary, and suggested follow-up questions.

    To ask a question in an exploration, launch AI Data Explorer from a data visualization or list or open an existing exploration. You will see a field with the placeholder "Ask Now Assist a question about data." For more information, see Launch AI Data Explorer.

    Note:
    • The question you ask has to be about data in one of the tables listed in the Query Generation Semantic Table Configuration table. These tables can include database views or Workflow Data Fabric tables. For more information, see Add a table to the semantic data layer.
    • If the data is from a protected application scope, access to that scope must be configured for AI Data Explorer. For more information, see Enabling access to protected scope applications for AI Data Explorer and Query Generation.
    • When you have submitted a question, you cannot submit another question or do other work in the exploration until your question is processed. You can cancel the processing of your question.

    When you write a question in an exploration, the AI converts the question to a database query and returns a response. The response includes the following sections:


    The response returned from a question to AI Data Explorer, showing the summary, data visualization, and suggested follow-up questions.

    Viewing the response source

    After you receive a response from the ServiceNow AI Platform, point at the response to see the technical details of the response. The source details include the following information:
    • The source table
    • The filter conditions
    • The metric
    • Any grouping criteria
    If the exploration is too narrow on the screen, select View source instead of pointing at the response.
    Source details for a response in an exploration that features table data.

    Tips for asking questions

    The goal of AI Data Explorer is to understand your prompts in your own words, delivering the analytics insights you want. However, if you do not know where to begin to formulate questions, or you're unsatisfied with the results, here are some tips:

    Name your table
    If you know the name of the table that contains the data you are interested in, add it to your prompt. Partial names or similar names are fine too.

    Example: Instead of "How many P1s were opened this week,” write "How many P1 requests were opened this week," which references the request tables. Better yet, write "How many P1 catalog requests were opened this week," which references the specific Catalog Requests table.

    Explain what you mean
    Query Generation tries to understand your terms, but you can add details to help guide it. If you get unexpected results, try being more specific about what you're looking for.

    Example: Instead of "Show me all stale incidents," write "Show me all incidents not updated in 5+ days."

    Be specific with names
    When filtering by referenced records like users, groups, or services, try to use their full display names for best results. The AI model may learn from previous queries in the same document, but using full names ensures accuracy.

    Example: Instead of "Cases with Workplace Ops," write "Cases with Workplace Operations."

    Edit and refine queries
    If the generated query isn't quite right, you can manually edit the filter conditions. The AI model will learn from your edits and apply them to future questions in the same document. For more information, see Change the parameters for a data source in an exploration

    Example: You ask "Show me critical incidents from the network team" but are not satisfied with the response. Instead of asking repeated variations of the same question, hoping for a better result, edit the filter to find records where Assignment Group is ‘Network Operations’ and Priority is ‘1 - Critical’. Then ask "Show me the inflow trend for these incidents over time”.

    Don't leave bad queries in your exploration
    The AI model uses the previous document context to write the next query. Therefore, if you cannot refine a query to get a useful response, delete it. Otherwise bad queries can accumulate in your exploration, leading to ever-worsening responses.
    Import complex filters
    For complex data that's hard to describe, import data visualizations or lists into your exploration. If the visualization or list is on a dashboard, you can apply any filters on the dashboard before importing. The AI model will use imported queries to understand related questions in the same document.

    Example: Don't ask "Show me servers about to retire by location." Such a prompt is vague and complex. Instead, import a visualization from a dashboard titled "PostgreSQL servers nearing retirement,” with the desired values for the dashboard filters Lifecycle State and Days Until Retirement pre-applied. Then ask "Show me the same servers but grouped by location”.

    Once you have a productive exploration going, with a lot of context, you may find that you can ask more abstract questions and get useful answers. However, these tips might help you get started.