Skip to navigation
InteractAsk questions of data with AI

Ask natural language queries with Sigma Assistant

View as Markdown

Sigma Assistant is a natural language query (NLQ) interface that allows you to ask questions about your data and interact with AI-generated responses. Extract factual information from your data, identify the drivers behind a metric, detect anomalies and get guided next steps, segment your data into meaningful groups, or ask questions that generate charts you can explore in a workbook. If a workbook you can access already exists that can answer your question, Assistant includes the workbook in its response.

The use of AI features is subject to the following disclaimer.

Sigma Assistant can answer questions about any data source that you have access to. By default, Sigma Assistant answers questions using data sources selected by your admin, but you can select another data source that fits your analysis before asking your question. Assistant uses multi-step reasoning to handle comparisons, trends, rankings, forecasts, and time-based or geographic questions, and supports conversations, so you can ask follow-up questions and iterate on a specific data analysis task without losing context.

System and user requirements

The ability to use Sigma Assistant requires the following:

  • An AI provider must be set up for your organization. See Set up an AI provider.
  • You must be assigned an account type with the Use Assistant permission enabled. See Account type and license overview for details.
  • To move analysis from Sigma Assistant into a workbook, you must be assigned an account type with the Create, edit, and publish workbooks permission enabled. See Account type and license overview for details.
  • To use warehouse tables as data sources in Assistant, you must have the View connections permission enabled on your account type.
  • To use data models as data sources in Assistant, you must have the View data models permission enabled on your account type.

Limitations

  • You cannot select a tagged version of a data model as a source in Sigma Assistant. If a data model is configured or selected by a user, only the published version of that data model is used to answer questions, and only users who have access to the published version of the data model can get answers from Sigma Assistant that reference that data model.
  • You can’t switch to a different agent within an existing chat conversation. To ask Assistant or a different warehouse agent a question, start a new chat.

Ask natural language queries with Sigma Assistant

To use Sigma Assistant to ask questions about your data:

  1. From Sigma Home, click Chat.

  2. In the Ask me anything about your data box, enter your question.

    Assistant interprets your question, searches data sources, and selects what it determines to be the best source to answer your question. If Assistant is set up to use only specific sources, Assistant searches only the admin-selected sources. Assistant uses semantic relevance, data source metadata, and data source usage to select the best data source from the data sources you have access to. You can optionally select a specific data source.

    Assistant also searches the workbooks you have access to for one that might already answer your question, so you don’t have to start a new analysis from scratch. For broader search across other document types, such as data models, see Search for files in your organization.

  3. If your question is vague, Assistant might prompt you to clarify what you’re asking for before responding.

    In response to a user question to tell me about sales, clarifying prompt from Assistant: Which sales view do you want? with 4 labeled options: Revenue trend (recommended), Bookings trend, Order volume, and Top customers with Skip and Continue buttons.

  4. Sigma Assistant provides a response, along with a summary of its reasoning and how long it worked for.

    Assistant response to Tell me about sales, with a Worked for 30s summary and a written answer stating sales revenue is trending up, from about 15.3M in 2021-06 to a peak of 75.1M in 2025-05, followed by a table of recent monthly revenue.

    For more details, see Review and interact with a response.

  5. You can continue to ask follow-up questions to refine your analysis or understanding of the data source. Assistant also suggests possible follow-up questions to explore in your analysis. If Assistant is still responding to your previous message, your follow-up message is queued. For example, you might follow up and ask “Please break down sales by region”:

    The end of Assistant's previous response, the follow-up question "yes, please break it down by region," and a new response with a bar chart showing revenue by region, with West leading at 219,700,751.80.

  6. If you want to save the full conversation, in the chat header, select Export as PNG to download an image of the conversation.

  7. If you think of an unrelated analysis that you want to perform in a different chat session, start a new chat:

    • If chat history is turned on for your organization, in the sidebar, click Chat to start a new chat. At any time, you can return to a previous conversation from the list.
    • If chat history is turned off for your organization, click Reset () to start a new conversation without leaving Sigma Assistant.

Review and interact with a response

After Assistant responds, you can review the steps Assistant performed to provide an answer, interact with the response by saving content for later, or provide feedback:

  • If Assistant ran a query to answer your question, you can review the query details for full transparency into the filters, calculations, and data sources behind the response:

    1. Click Worked for… to expand the details of the actions Assistant performed.

    2. In the details, click Gathered data from… to expand the queries that Assistant performed as part of its analysis.

    3. Select a data source to review the calculated output in a summary panel.

    4. Click View SQL to review the exact query Assistant sent to your data platform.

      Assistant response to the question, with the working details expanded and the results table and SQL query shown on a summary panel.

  • If the response includes a chart, you can click Download chart as PNG to save an image of the chart to your computer.

    A Revenue Region bar chart with the cursor hovering over the download icon, showing a Download chart as PNG tooltip.

    The Download permission is not required to download a PNG of a chart.

  • If you’re assigned an account type with the Create, edit, and publish workbooks permission enabled, hover over a chart and select Open in workbook to open an exploration in a new tab with the selected output and any underlying data tables.

  • Click (Copy response) to copy the response to share with someone else.

  • To provide feedback to Sigma about the quality of the response from Assistant, select (Good response) or (Bad response). If you select Bad response, optionally provide details about what was wrong. Only the feedback details are shared with Sigma. Details about your conversation with Assistant are not provided to Sigma.

    Admins can also review scores and feedback for Sigma Assistant interactions in the AI usage table. See AI usage dashboard.

Review a summary of source details and your conversation

At any point in your conversation, you can select (Show summary panel) to display a summary of your conversation:

  • Review the sources used by Assistant throughout your conversation. Select a source to preview the data and confirm it is the relevant source. When previewing a source, click Open table or Open data model () to open the source in a new tab.
  • Review your past prompts in the conversation. Select a prompt to review the response to that prompt in the summary panel. For example, you can compare different responses side by side from your analysis.

The summary panel open beside a conversation with Assistant, listing F_SALES and PLUGS_ELECTRONICS_HANDS_ON_LAB_DATA as sources and two past prompts, tell me about sales and yes, please break it down by region.

Search and explore data sources

Sigma Assistant has access to all data sources that you can access in your organization by default, unless an admin has selected specific sources for Assistant to use first. If you know exactly which data source you want to ask questions of, you can select it before asking questions with Sigma Assistant:

  1. From Sigma Home, click Chat.

  2. Select +, then search for one of the following data sources that you can access:

    • Table in your data platform
    • Published version of a data model, or a specific element within it

    Sigma returns all matching tables and data models that you have access to use. If your admin has selected specific sources for Assistant, those sources include an indicator.

    All data access grants that apply elsewhere in Sigma also apply in Sigma Assistant. For example, if row-level security or column-level security is applied to a data source, Sigma Assistant only works with data that you have access to. For a data model, you must have access to the published version.

    Your admin can select sources to be used by Sigma Assistant by default.

Interact with warehouse agents in Sigma Assistant

When you ask questions of Sigma Assistant, you can choose a Snowflake Cortex Agent or a Databricks Genie Agent in your organization to answer the questions. For setup steps, see Use warehouse agents with Sigma.

Requirements

If you chat with a warehouse agent instead of Assistant, an AI provider is not required.

  • Using warehouse agents for Assistant must be turned on in the AI settings for your organization.
  • You must be granted Can use access to the warehouse agent, schema, database or catalog, or connection that contains the agent.
  • If you connect to Snowflake using OAuth and want to use a Snowflake Cortex Agent, your role must be granted USAGE privileges on the agent in Snowflake.
  • If you connect to Databricks using OAuth and want to use a Databricks Genie Agent, your role must have CAN VIEW or CAN RUN on the Genie Agent in Databricks.

Each warehouse agent runs using the role configured in the connection to the data platform. For details, see Use warehouse agents with Sigma.

Ask natural language queries of a Snowflake Cortex Agent

In Sigma Assistant, ask natural language queries using a specific Snowflake Cortex Agent:

  1. From Sigma Home, click Chat.

  2. In the agent dropdown, choose a warehouse agent that you have access to. Sigma Assistant is selected by default.

  3. In the Ask me anything about your data box, enter your question.

  4. Sigma Assistant provides a response.

    You can continue interacting with Sigma Assistant. See Ask natural language queries with Sigma Assistant.

Limitations with warehouse agents

When using a Snowflake Cortex Agent or Databricks Genie Agent with Sigma Assistant, the following limitations apply:

  • Asking questions about a data source is not available.
  • Questions about data modeled in Sigma, such as a data model defined in Sigma, or a metric defined on a table in Sigma, cannot be accurately answered. A warehouse agent has access only to the data in your data platform.
  • Databricks Genie Agents cannot create charts.

Manage chat conversations

If chat history is turned on for your organization, all chat conversations that you have with Assistant or a warehouse agent are stored in Sigma. Chats are available until the configured retention period for your organization is reached.

Your chat history is only visible to you, and admins can review all conversation details, including deleted chats, in the AI usage dashboard if the AI usage dashboard is configured.

To revisit, rename, or delete a previous chat conversation:

  1. In the Chat section of Home, review the automatically named chats.
  2. To locate an older chat, scroll the list until you locate it.
  3. Select a chat to send new messages.
  4. To rename a chat, select More (), then select Rename to enter a new name for the chat.
  5. To delete a chat, select More (), then select Delete. The chat is deleted from your history, but admins can still review it in the AI usage dashboard.