8.1 AI/BI Dashboards Overview & Canvas Layout

8.1 AI/BI Dashboards Overview & Canvas Layout

Exam Focus: Databricks AI/BI Dashboards (formerly code-named Lakeview) represent the modern, AI-native dashboarding framework built directly on top of Unity Catalog and Databricks SQL Warehouses. Candidates must master the architectural decoupling of Draft vs. Published states, Unity Catalog security inheritance, execution credential modes (Run as Viewer vs. Run as Owner), and multi-page canvas layout optimization.

Introduction to Databricks AI/BI Dashboards

Databricks AI/BI Dashboards are native visualization assets designed to deliver fast, governed analytics across enterprise datasets stored in Delta Lake. Historically, organizations using Databricks relied on legacy Databricks SQL (Redash-based) dashboards or external third-party BI tools (such as PowerBI, Tableau, or Qlik). Traditional external BI tools require exporting bulk data into proprietary intermediate extracts or maintaining complex external semantic layers. In contrast, AI/BI Dashboards execute queries directly against Databricks SQL Warehouses (Serverless or Pro), eliminating data extraction overhead and ensuring zero latency between Delta Lake updates and visual reporting.

Key architectural pillars of Databricks AI/BI Dashboards include:

• Unity Catalog Native Governance: All dashboard queries inherit table permissions, row filters, column masks, and audit logging defined in Unity Catalog.
• No Client Data Extract Storage: Visualizations render directly from live query execution results returned by Databricks SQL Warehouses.
• AI-Powered Semantic Integration: Dashboard datasets easily convert into AI/BI Genie spaces, enabling natural language data exploration for business users.
• Modern Canvas Layout System: A responsive, grid-based layout supporting multi-page navigation and dynamic parameter binding.

---

Core Architecture: Draft vs. Published States

A central operational concept on the Databricks Data Analyst certification exam is the strict separation between Draft and Published states within AI/BI Dashboards. This design allows data analysts to continuously iterate on complex data models, modify SQL query logic, and redesign canvas layouts without disrupting business stakeholders who rely on live production reports.

When an author opens an AI/BI Dashboard, they enter the Draft workspace environment. Edits made to dataset SQL definitions, widget placements, text annotations, or color palettes are immediately saved to the draft version. To make these updates visible to end-users, the author must explicitly click Publish. Databricks creates an immutable snapshot of the dashboard assets (queries, layout configuration, parameter definitions). If an author subsequently makes further experimental edits in Draft mode, end consumers viewing the Published version continue to see the stable snapshot until the next explicit publication event.

---

Security & Execution Credential Modes

Databricks AI/BI Dashboards provide granular access control by combining workspace file permissions (CAN VIEW, CAN EDIT, CAN MANAGE) with Unity Catalog data access privileges. When publishing a dashboard, the author must configure one of two critical Credential Execution Modes:

1. Run as Viewer

In Run as Viewer mode, dashboard queries execute using the security context and identity of the individual end-user currently viewing the dashboard.

• Security Boundary: Unity Catalog evaluates table grants (SELECT), row filters, and column masking rules specifically for that viewing user.
• Data Isolation: If User A has access to region = 'North America' via a row filter, while User B has access to region = 'Europe', viewing the exact same dashboard produces distinct, isolated data results for each user.
• Requirement: Every viewer must have SELECT privileges on all underlying Delta tables and active compute access to the designated Databricks SQL Warehouse.

2. Run as Owner

In Run as Owner mode, dashboard queries execute using the embedded identity and credentials of the dashboard publisher (or designated owner service principal).

• Security Boundary: Viewers interact with the pre-aggregated metrics rendered on the dashboard without needing direct access to the raw underlying Delta tables in Unity Catalog.
• Use Case: Ideal for distributing executive summaries or high-level KPI scorecards derived from sensitive raw PII datasets. End users see the final visualizations, but Unity Catalog prevents them from running ad-hoc SQL queries against the raw source tables.

-- Example Unity Catalog Row Filter enforced under 'Run as Viewer' mode
CREATE FUNCTION main.security.region_filter(region_col STRING)
RETURN IS_ACCOUNT_GROUP_MEMBER('executives') OR region_col = CURRENT_USER();

---

Canvas Layout & Widget System

The AI/BI Dashboard canvas uses a responsive, grid-based layout engine designed for multi-device viewing. Analysts organize content across three primary widget types:

1. Visualization Widgets: Render graphical charts (bar, line, scatter, pivot) generated from underlying SQL datasets.
2. Text & Markdown Widgets: Provide contextual narrative, section headers, analytical callouts, operational documentation, and dynamic formula explanations using full Markdown styling.
3. Filter & Parameter Widgets: Provide interactive UI controls (text boxes, dropdown pickers, date range sliders) bound directly to dataset parameters or categorical table columns.

Multi-Page Organization

AI/BI Dashboards support multi-page canvas structures, allowing analysts to create up to 15 distinct pages per dashboard asset. Pages appear as horizontal tabs across the top of the canvas layout. Multi-page organization allows separating reporting domains—such as dedicated pages for Executive KPI Overview, Regional Sales Deep-Dive, and Inventory Logistics—within a single, unified governed asset.

---

Datasets & AI/BI Genie Space Integration

Unlike legacy dashboards where each visual widget required a separate, isolated SQL query, AI/BI Dashboards introduce centralized Dashboard Datasets. An analyst defines SQL queries in the dataset tab, and multiple visualization widgets across various canvas pages can consume and slice data from those shared dataset queries. Furthermore, Databricks enables seamless synergy between AI/BI Dashboards and AI/BI Genie. Authors can convert defined dashboard datasets into an AI/BI Genie space with a single click, empowering business users to transition from viewing static canvas charts to asking natural language questions against the exact same governed metrics.