Skip to content

AI Analyst

AI Analyst is a conversational, AI-powered analytics feature in Sprinkle Data that enables users to explore their data and generate insights by asking questions in natural language.

Instead of manually creating reports or writing SQL queries, users can ask business questions and receive data-driven answers through visualizations, tables, and contextual explanations.

AI Analyst uses a configured Large Language Model (LLM) and a governed Domain Model to understand business terminology, identify relevant data, and generate queries based on the available data and defined business context.

The Domain Model acts as the foundation of AI Analyst by defining the context and scope of data the assistant can use. Administrators can configure the LLM, generate the Domain Model, and control which models and tables are available for analysis.

Once configured, users can interact with AI Analyst to explore their authorized data, investigate trends, compare metrics, and uncover actionable insights through a conversational interface.

Before configuring AI Analyst, ensure that you have:

  • Access to the Admin section in Sprinkle Data.
  • A valid API key for the LLM provider you intend to use.
  • The required data models and tables available in your Sprinkle Data environment.
  • An understanding of which data models and tables should be accessible to AI Analyst.

The first step is to configure the Large Language Model (LLM) that AI Analyst will use to interpret questions and generate analytical responses.

  1. Log in to Sprinkle Data.
  2. Navigate to Admin โ†’ AI Settings.
  1. Click Configure New LLM.

  2. Select the required LLM provider from the available options.

  3. Enter the following details:

    • Name: A descriptive name to identify the LLM configuration.
    • Model Name: The name or identifier of the model you want to use.
    • API Key: The valid API key required to connect to the selected LLM provider.
  4. Review the entered details and click Add.

The LLM configuration is now added and available for use in the AI Analyst setup, subject to the configuration being valid and supported.

The Domain Model provides the curated business and data context that AI Analyst uses to answer questions about your data warehouse.

It scans relevant information from your tables, dashboards, reports, and team activity to identify important business context. Using the configured LLM, it generates descriptions for tables, columns, models, and fields, along with suggested prompts and example queries.

The generated context helps AI Analyst understand your data structure, interpret business questions, and identify the relevant data for query generation.

When you generate a new Domain Model version, Sprinkle Data scans the available context and produces an updated version. Promoting a version makes it the live context used by the chat assistant.

  1. Navigate to the Domain Model tab in Sprinkle Data.
  2. Open the configuration workflow to generate a new Domain Model.

In the configuration modal, specify the context and data scope that AI Analyst should use.

Enter notes to provide additional business context for the LLM.

These notes can help clarify business terminology, explain how your organization interprets its data, and guide the generation of relevant descriptions and example queries.

For example, you might explain what your organization considers a qualified lead or how a particular revenue metric should be interpreted.

Select the configuration options that determine which parts of the Domain Model the LLM is allowed to configure.

Review the available options and select the ones appropriate for your requirements.

Configure the scope of data that AI Analyst is allowed to access during analysis.

The available controls include:

  • Models: Select the data models that AI Analyst can use.
  • Allowed tables: Specify the tables that can be accessed for analysis.
  • Restricted tables: Identify tables that must be excluded from the configured scope.

Review the selected models and table permissions carefully before proceeding. AI Analyst should only be configured with the data required for its intended use cases.

The selected scope determines which data is available to AI Analyst. It does not grant unrestricted access to the entire warehouse.

After configuring the notes, LLM options, and data scope:

  1. Review the configuration.
  2. Click Run.
  3. Allow Sprinkle Data to scan the selected context and generate the Domain Model.

Once the process completes, the generated Domain Model is available for review and version management.

Important: Generating a new version and making it live may be separate steps. If the interface requires you to promote the generated version, promote it before using it as the active context for AI Analyst.

Once the LLM and Domain Model are configured and the appropriate Domain Model version is active, you can start asking questions about your data.

Navigate to the AI Analyst screen in Sprinkle Data.

Enter a question in natural language about the data available within your configured scope.

For example:

  • What were our total sales last month?
  • Which region generated the highest revenue?
  • Show me the trend in customer acquisition.
  • Compare this monthโ€™s performance with the previous month.

AI Analyst interprets your question using the configured Domain Model, generates and executes the relevant query, and presents the results.

Review the response and any available charts, tables, or supporting explanations.

Ask follow-up questions to investigate the results further.

For example, after asking which region generated the highest revenue, you can ask which products contributed most to that regionโ€™s performance.

AI Analyst operates within the data scope configured during Domain Model creation.

It can use the models and tables included in the configured scope to answer questions. Data outside that scope is not available to the assistant through this configuration.

For example, if the Domain Model includes sales and customer tables but excludes employee records, AI Analyst can answer questions using the permitted sales and customer data but will not use the excluded employee records.

Important considerations:

  • Include only the models and tables required for your intended analytical use cases.
  • Exclude sensitive or irrelevant tables from the configured scope.
  • Review the Domain Model whenever your data structure or business requirements change.
  • Ensure that any additional user-level permissions and row-level security policies are configured appropriately. Domain Model scope should not be treated as a replacement for other access-control mechanisms.

The complete setup follows this sequence:

  1. Configure an LLM under Admin โ†’ AI Settings.
  2. Open the Domain Model tab.
  3. Add Domain Model notes and select the LLM configuration options.
  4. Define the accessible models and allowed or restricted tables.
  5. Click Run to generate the Domain Model.
  6. Promote the generated version if required.
  7. Open AI Analyst and start asking questions about your data.

After completing these steps, AI Analyst is ready to support conversational data exploration within the configured scope.