Use AI Assistant

The AI assistant adds a conversational layer to GoodData, letting you ask questions in plain language and get answers that align with the metrics, datasets, and business terms already defined in your semantic model - no knowledge of SQL or MAQL is required.

The assistant can also search existing dashboards, visualizations, and metrics, create new visuals on the fly, refine its output based on follow‑up instructions, and even field general questions about your data.

Every query is interpreted through the semantic layer and sent to the language model as metadata only, so raw data never leaves your environment. Responses are fully traceable since the exact metrics and filters used in the answer are always listed.

For developers, we also support working with the assistant via our API, as well as our React and Python SDKs. You can start by reading our the introductory article on our blog, or head straight to GoodData.UI, Python SDK or API documentation.

The assistant is currently optimized to work with the English and Brazilian Portugese languages. Support for other languages is planned for future releases.

This feature is still in active development and we are aware of certain limitations.

Access AI Assistant

Before you can chat, an administrator must first configure the assistant. Once configured, the assistant is available in any workspace where you have the AI assistant user permission.

Open a workspace and click the assistant icon in the upper‑right corner.

Screenshot of the top right portion of a workspace, highlighting the AI assistant button.

A chat panel slides in on the right.

Screenshot of a newly opened, empty ai assistant dialog.

Hit the expand icon to stretch the panel for longer conversations.

Close up screenshot of the assistant highlighting the expand button in top right corner of the assistant.

Ask Business Questions

Type a plain language question such as “What was our total profit last year?”. The assistant translates the request into a metric query and returns the answer as a visualization - in this case, a headline:

Screenshot of the assistant showing query 'What is the overall profit for last year?' and a response in form of a generated Headline showing the value of profits for previous year.

While you type, the Assistant offers autocomplete suggestions drawn from existing metrics, attributes, and facts. This helps you confirm field names and, when several similar objects exist, choose exactly the one you need:

Screenshot of the assistant showing the 'What was our total profit' typed in the input field and an autocompletion dialog showing profit related metrics popping up.

TIP

The clearer you make your metric names and descriptions, the better the assistant’s answers will be.

However, if multiple metrics, dimensions, or filters match a user’s question, the AI Assistant asks clarification questions to help users choose the correct object. This is an experimental feature that ensures accurate results even when similar or overlapping data objects exist.

Follow‑up questions stay in context. For example, “Break it down by quarters” produces a table with one row per quarter:

Screenshot of the assistant showing query 'Break it down by quarters' and a response in form of a table with 4 rows, each showing profit for a quarter.

The assistant tries to select a type of visualization that best fits your query.

Save Visualizations

The assistant can create the following types of visualizations:

You may specify the type directly, for example by asking it to “Show it as a bar chart.”

Screenshot of the assistant showing query 'Show it to me as a bar chart' and a response being a bar graph with 4 bars, one for each quarter's profit.

For attributes, the generated visualizations use default labels.

Hover over the visualization and click the save icon in its upper‑right corner.

Same screenshot of the assistant again, this time showing the mouse cursor hovering over the visualization and the save button showing up.

After naming it, the chart is saved to the workspace and is ready for tweaks in Analytical Designer or immediate placement on a dashboard.

Need a fresh start? Click the reset icon in the upper‑left corner of the chat to clear the conversation.

Same screenshot of the assistant again, this time showing the mouse cursor hovering over the top left corner of the assistant, over the reset button.

Unless reset, the assistant remembers your conversation in each workspace separately.

Search Workspace

You can use the assistant to search the workspace for:

  • Dashboards
  • Visualizations
  • Metrics
  • Facts
  • Attributes

For example, “Show me profit‑related analytics in this workspace.” lists matching objects as clickable links:

Screenshot of the assistant showing query 'What profit related analytics are there in this workspace?' and the response being a list of metrics, visualization and facts, each showing up as a clickable link.

Attribute Filtering in AI Assistant

The AI Assistant can apply filters from natural language, even when you do not type exact attribute names or values. It finds the best matching value in your data and applies it automatically. This makes asking questions easier and more reliable for everyday use.

The AI Assistant understands these questions by matching values in your data. No data is shared with any LLM to enable this behavior.

How It Works:

When your question includes a value that looks like a filter, the AI Assistant searches available attributes and their values, finds the best match, and applies it to the answer or visualization.

You do not need to know the exact attribute name or value.

Examples:

  • You provide an attribute name and an approximate value

    If you ask:
    What are the total sales for product category computers?

    The AI Assistant:

    1. identifies the attribute Product Category and value Computers
    2. searches for the closest matching value
    3. applies the filter, for example:
      Product Category = Computers and Accessories

  • You provide only a value, without naming the attribute

    If you ask:
    What are the total sales for computers?

    The AI Assistant:

    1. treats computers as a filter value
    2. finds the attribute that best matches it
    3. applies the filter, for example:
      Product Category = Computers and Accessories

    Another example:
    What is my revenue coming from California?
    is interpreted as:
    State = California

  • No matching value is found

    If you ask:
    What are the total sales for kitchen appliances?

    And no matching value exists in your data, the AI Assistant responds with a message such as:
    I couldn’t find a matching value for kitchen appliances. Try refining your question or check your data.

Configure AI Assistant