What AI can understand dashboards?

Summary

• Genie AI understands dashboards by combining semantic understanding of business context (metrics, periods, roles) with natural language query capabilities.
• Genie Agents are built automatically from your dashboard datasets and displayed as an "Ask Genie" button, letting viewers ask follow-up questions in plain English.
• The system uses conversational interaction: asking for clarification when questions are ambiguous, maintaining context across follow-up questions, and suggesting visualizations.
• Genie Agents point to Unity Catalog metric views as the semantic layer, defining dimensions, measures, synonyms, and business metadata for accurate answers.
• Governance is integrated at the data layer: row-level and column-level access controls from Unity Catalog enforce automatically across all dashboard queries.

What AI can understand dashboards?
Genie AI understands dashboards through semantic awareness of your business logic, not just keyword matching. Genie comprehends what specific metrics mean in your organization, which fiscal period a reference like "Q3" implies, which data matters for each user's role, and how to connect dashboard questions to underlying data. When a business user asks "How did we perform against targets this quarter?" Genie retrieves the right tables, applies the correct definitions, and respects access controls. All of this occurs through natural conversation.

How Genie works with dashboards

Every published AI/BI dashboard includes an automatic "Ask Genie" button. When clicked, a companion Genie Agent opens alongside the dashboard. This agent:

• Answers natural language questions about the dashboard's data without requiring SQL or dashboard knowledge
• Generates appropriate visualizations automatically (selecting chart types, aggregations, and drill-down paths based on the question)
• Maintains context across multiple follow-up questions in a single conversation
• Seeks clarification when questions are ambiguous, rather than guessing and returning wrong answers
• Suggests related questions to help users explore data beyond fixed dashboard views

This companion agent is built automatically from your dashboard's datasets and visualizations, so it understands the shape and purpose of your existing dashboard.

What makes Genie's understanding semantic

Capability	What it enables
Business context awareness	Genie knows what "targets" means for your organization, which fiscal period "Q3" refers to, and which metrics apply to each role.
Metric and dimension definitions	When queries reference metrics or dimensions, Genie uses defined synonyms and business metadata (e.g., "revenue" = "total transaction value").
Trusted logic and validation	Genie can integrate validated formulas and calculations from Unity Catalog, ensuring answers follow your governance rules.
Multi-step reasoning	Complex questions like "Compare Q3 revenue growth to the seasonal average" are decomposed into logical steps using the data model.
Contextual suggestions	Based on the user's question and role, Genie suggests related metrics or drill-down paths that are likely valuable.

Configuration for accuracy and governance

Data teams configure Genie Agents using:

• Instructions: Provide context about your business, define ambiguous terms, and set guardrails for what Genie should and shouldn't attempt to answer
• Trusted assets: Point Genie to Unity Catalog metric views (the semantic layer) rather than raw tables. Metric views define dimensions, measures, synonyms, and business metadata, giving Genie the context needed for accurate answers
• Quality monitoring: Track which questions Genie answers well and which need refinement, then adjust instructions or training data
• Confidence voting: Users can indicate whether Genie's answers are correct, informing future improvements

All governance policies from Unity Catalog apply automatically: row-level restrictions, column masking, and data classification flow through to every Genie query on the dashboard.

Use cases

Finance teams exploring budget variances. An analyst clicks "Ask Genie" on a financial dashboard and asks "Which departments are over budget this quarter?" Genie queries the underlying data warehouse, applies the user's row-level access (so they only see their division's data), and returns a dynamic table with drill-down links to supporting details.
Operations monitoring dashboards. A plant manager asks Genie "Which production lines had the most downtime this week?" Genie understands that "downtime" maps to a specific calculation in your data model, retrieves the relevant data, and suggests follow-up questions like "What was the mean time between failures for each line?"
Sales performance conversations. A sales director uses the Ask Genie feature to explore "How many deals closed in EMEA this quarter in recent quarters?" Genie maintains context through follow-ups: "What's the pipeline for Q4?" and "Which vertical is driving that growth?" without losing the initial geographic and time filter.
Data validation and quality checks. Data analysts use Genie on a data quality dashboard to ask "Why did record count drop yesterday?" Genie connects the question to your metadata, recent ingestion patterns, and historical baselines to narrow down likely causes.

FAQs

Can Genie understand any dashboard, or only specific ones?

Genie automatically understands dashboards built in Databricks AI/BI. The system analyzes your dashboard's datasets, visualizations, and the underlying metric definitions in Unity Catalog. For dashboards external to Databricks, you can still use Genie Agents to query the same data by building a separate agent and pointing it to the same metric views.

What if my dashboard has incomplete or poor metadata?

Genie works effectively with complete business context. If metric definitions, synonyms, or relationships are missing, provide this information in Genie Agent instructions or in Unity Catalog metric views. Incomplete metadata may cause Genie to misinterpret questions or suggest less relevant answers. Investing in metadata quality directly improves Genie's accuracy.

How does Genie respect my data access controls?

Every Genie query applies Unity Catalog policies automatically. If you're restricted from viewing a table or column, Genie won't expose that data. Row-level access controls filter results so each user only sees data they're authorized to access. Audit logs capture every Genie query, so compliance teams have visibility.

Can Genie answer questions that aren't directly in my dashboard data?

Genie can access any data in Unity Catalog that your dashboard is authorized to query. If you ask about a related table that isn't visualized in the dashboard, Genie can still answer as long as that table is in the same catalog and you have access rights. This makes Genie valuable for exploring data beyond fixed dashboard views.

How long does it take to set up Genie for a dashboard?

Genie is generated automatically from your dashboard, so it's ready to use immediately upon publication. Fine-tuning (adding instructions, improving definitions, or incorporating trusted logic) typically takes a few hours per dashboard depending on complexity.