Use a Genie space to explore business data

This article explains how to test, use, and provide feedback on Genie spaces, a no-code self-service chat interface where you can ask questions about your company’s data.

Overview

AI/BI Genie provides a conversational interface for querying your data using natural language. It uses generative AI tailored to your organization’s business terminology and data and can incorporate user feedback to refine responses over time.

Genie offers the following key capabilities:

Enables users to ask natural language questions to query enterprise data directly.
Provides relevant responses derived directly from organizational data.
Supports self-service to address questions beyond the scope of dashboards and reports, without requiring advanced tools or expertise.

The following image identifies parts of the Genie space UI:

A labeled identifying parts of the Genie space UI. Each number is paired with an explanation that follows the image.

Chat window: Use the chat window to ask questions and read Genie’s responses. It is open by default when you open a Genie space.
Threaded chat history: Your interactions with Genie are saved, and you can access them from the threaded chat history to the left of the chat window. Each chat thread maintains context about previous questions it has been asked. You can use new threads to organize your Genie space by topic.
Data tab: Like dashboards, Genie spaces are based on a subset of your company’s data. The data tab shows the tables selected for this Genie space. Expand each table to view column names and descriptions. Click Open in Catalog Explorer to view table details from the Catalog Explorer UI. To learn more about the information you can access in Catalog Explorer, see What is Catalog Explorer?.
History tab: The history tab shows individual Genie interactions, not threaded by topic. When you converse with Genie, you can rate the response or request a review from someone with elevated permission on the space. Quality ratings and review correspondence appear in the history tab.

Use a new Genie space

Genie spaces are configured by domain experts, such as data analysts familiar with SQL. The setup process includes:

Defining the purpose of the space.
Selecting and annotating relevant data.
Adding plain text instructions to help Genie understand business-specific jargon.
Providing example SQL queries for Genie to learn from.
Incorporating existing SQL functions for business critical or complex calculations.

Note

This article assumes that a Genie space has already been created for you. See Set up and manage an AI/BI Genie space for more information.

Space authors typically start with basic instructions and then refine the space as they learn from user interactions. As users ask questions and introduce specific terminology or jargon, authors update the space’s instructions to improve how Genie interprets and responds to queries. This iterative process helps the Genie space evolve and provide more accurate answers based on user feedback. It can be helpful to think of Genie as a new employee who improves with clear guidance and regular input.

Chat with Genie

Most Genie interactions take place in the chat window. Click Chats to open the chat window anytime. The first time you open a Genie space, your chat history is blank, and the chat window shows the following components:

Title and description: This shows the title and description specified by the author.
Example questions: Example questions are typically added by the space editor to capture common inquiries and illustrate the types of questions the space is designed to answer. Click an example question to start a new chat.

Note

The Explain dataset button is included as an example question by default. When you click that button, Genie uses available table metadata and defined primary and foreign key relationships to explain the tables and how they’re connected.
Question text field: Enter a new question by typing into the question field near the bottom of the space. When Genie generates a response, it appears above the question input field. You can rate response quality and request a manual review of any response from this part of the Genie UI. See Review a response.

Start a new chat

Click an example question or enter a new question to start a conversation. Each conversation thread maintains its context, letting you build on previous questions. For example, if you ask, “Show me sales by product line,” you can refine results in the same thread by asking, “Only for July 2024.”

Note

Context from earlier conversations in a thread is used to inform later conversations in the same thread. This context does not carry over to other chats.

Click + New chat in the threaded chat history to start a new chat. By default, the chat is named with the initial prompt.

What questions can I ask?

A Genie space is designed to return query results and data insights that directly answer your questions. It does not interpret results or provide recommendations. For the best outcomes, ask specific, precise questions based on the available data, rather than seeking explanations or advice.

The table below outlines example question types Genie can answer and non-examples of questions it cannot process effectively.

Questions Genie can answer	Questions Genie cannot answer
What were the sales this quarter?	Why were sales low this quarter?
Which customers generated the most revenue?	How can I get more revenue from Customer M?
Which sales representatives generated the most revenue?	What incentives should I offer my sales team to generate more revenue?
What were the top-performing marketing campaigns in 2024?	How can I improve my marketing strategy?
Which sales regions generated the most profit?	What are the best sales regions?

Define terms in a conversation

When you enter a question, Genie uses table and column metadata, instructions provided by space editors, and context from the current conversation thread to generate a response. If Genie can’t answer the question, it might respond by asking a follow-up clarification question. Or, it might explain that it can’t answer based on the provided information. You can continue to work with Genie to get an answer to your question by defining words, relationships, or business jargon in the conversation.

For example, if you ask something like, “How many customers churned in the past year?” Genie might need additional information, such as how churn is defined. It could respond with a clarifying question or state its limitations. You can then provide more context in the same conversation, which Genie uses to improve its understanding for future interactions in that thread.

Note

Genie, like other large language models (LLMs), can exhibit non-deterministic behaviors. This means that you might receive different outputs when submitting the same prompt multiple times.

Review a response

When Genie provides a response as a table of values or an individual value, it uses the provided context and instructions to process your question and generate a SQL statement that returns the answer. If you are familiar with SQL, you might want to inspect the generated query. To view the generated query, click Show generated code at the bottom-right of the response.

Generate visualizations with quick actions

A set of Quick actions tiles follow responses that include tabular data. You can use them to generate visualizations.

Quick action tiles that suggest different visualization options.

You can also generate a visualization by describing it in words after getting a query result. Download a visualization as a PNG file by clicking the download icon in the upper-right corner of the visualization.

Format results

To format a column in a table within a response, click the kebab menu in the column header, then click Format to view the available options. Formatting options depend on the column’s data type.

A numeric column with formatting options such as currency and percentage.

Download the result set

You can download up to approximately 1GB of results data as a CSV. The final file download size might be slightly more or less than 1GB, as the 1GB limit is applied to an earlier step than the final file download. To download results, click the download icon in the response.

Provide feedback

Authors have a holistic view of user interactions within the space. They can review questions and feedback and engage directly with users through the Genie space interface when a response is marked for review. Your input is crucial to understanding what users want from the space, so it’s important to use the built-in tools to provide feedback. Review responses critically, rate response quality, and flag interactions you need clarification on for review.

To provide feedback, use the buttons near the lower-left corner of the response to do one of the following:

To rate response quality: Click the thumbs up or thumbs down icons to rate a response as good or bad, respectively.
To request a review of a response: If you’re unsure of the response quality and want a space editor to review the interaction, click the Ask for review icon. You can include an additional question or comment for the space editor with your request.

Review chats

Your chat history stores a record of your threaded conversations and remains available until the space is deleted. Most users can only view their own history. Each chat thread maintains its own context, meaning early messages inform subsequent questions and responses only within that thread. Context established in one chat does not carry over to others. You can open past chats to refine your questions or insights, enabling you to build more complex analyses and gain deeper understanding over time.

To edit the name of a chat:

Hover over the conversation thread and click the edit icon.
Enter a new name into the text field. Press enter to save the new name.

To delete a chat:

Hover over the thread and click the trash icon.
Click Confirm and Delete Chat to remove the chat from your history.

Note

Users with elevated privileges on the space can view all chats. Deleting a chat from your chat history also deletes it from their history page.

To review a previous response:

Open the conversation by clicking its name.
Scroll through the history in the chat window.

Note

Query results persist for seven days. If the interaction you want to review is older than that, you can regenerate the response or rerun the generated SQL.
(Optional) Regenerate a response: Click the Regenerate icon to reprocess your original question.
(Optional) Rerun the SQL query: Click Show generated code. Then, click Run to rerun the query without having Genie reprocess the request.

Best practices for using a Genie space

The following list includes recommended best practices for using Genie spaces:

Ask one question at a time: Genie is best at answering clear and precise questions directly connected to the available data. To get the best results, ask a specific question. See What questions can I ask? for examples of what constitutes a good question. You can ask additional questions that build on the previous interactions. Organize your questions by using separate conversation threads for each topic.
Provide feedback on response quality: Genie spaces improve when users provide feedback. When you use the thumbs up or thumbs down buttons to rate the response quality, space creators can use that feedback to decide how to refine the space.
Request reviews and add comments: You can communicate asynchronously with the space author or editor by requesting and including a comment. This flags the response for the space authors, who can inspect the response and either confirm it is correct, or provide a correction if not.
Update queries with fresh data: If you have recurring questions for which you want to refresh the query results, use the threaded chat history to open the original response and the generated SQL. You can rerun the query without having to generate a new one.
Communicate context to space editors: If you find yourself explaining the same context multiple times in your conversations, consider contacting the space’s editors so that they can incorporate that context into the space’s instructions.
Carefully review responses: Genie can sometimes respond with hallucinations or errors, as with other LLM tools. Check the accuracy of the provided responses.