Guides
  1. Get Started

Getting Started with AI Insights

Learn how to enable and use BEEM AI Insights to ask natural-language questions about your data and get instant SQL-powered answers.

BEEM AI Insights is a conversational AI feature that lets you ask questions about your data using natural language instead of writing SQL manually.

Prerequisites

  • At least one deployed dataset in the Warehouse
  • Datasets marked as searchable by BEEM AI
  • Dataset descriptions (recommended for better results)

Enable Datasets for AI Search

  1. Navigate to the Warehouse in the left sidebar
  2. Open the folder containing your dataset
  3. Click on the specific dataset
  4. In the right panel, locate the Automation section
  5. Toggle ON "Mark dataset as searchable by BEEM AI"
  6. Click Save

Add Dataset Descriptions (Recommended)

  1. Open the dataset in the Warehouse
  2. Locate the Description field
  3. Use "Generate a description" for auto-generation based on columns
  4. Review and optionally use "Improve description" to refine
  5. Save changes

Open AI Insights

  1. Click AI Insights in the left sidebar
  2. The chat interface opens, showing the number of searchable datasets

Ask a Question

  1. Type a question in natural language:
    • "How many orders do I have?"
    • "What is the average order value?"
    • "Describe my datasets"
    • "Tell me what [dataset name] is used for"
  2. The AI generates a SQL query, runs it, and displays results

Work with Results

  1. View SQL: Click the code icon to see the generated query with syntax highlighting
  2. Copy Data: Transfer query results to your clipboard
  3. Copy Query: Extract the SQL for use elsewhere
  4. Export as TSV: Download results for further analysis

Continue the Conversation

  1. Ask follow-up questions in the same conversation for context
  2. Use suggested follow-up questions for deeper exploration
  3. Start a New Conversation for a different topic
  4. Access past conversations from the history panel

How It Works Under the Hood

  • Built on AWS Bedrock (multi-model: Anthropic, Mistral, Cohere)
  • Operates within your dedicated cloud environment. Your data never leaves your infrastructure.
  • Workspace-scoped: AI can only access data within your current workspace
  • Works with both owned and shared datasets
  • Chats are not used to train AI models

Best Practices

  • Be specific and include dataset names when possible
  • Add comprehensive descriptions to all datasets you want to query
  • Start with simple questions before complex analysis
  • Review generated SQL queries to verify accuracy
  • Use suggested follow-up questions for exploration
  • Verify results before using them for critical decisions

Need help? Contact [email protected].

Updated 4 days ago