AI/BI Dashboards
Databricks · Databricks SQL & BI
Theory
AI/BI dashboards (formerly Lakeview dashboards, which replaced the legacy Databricks SQL dashboards) are the built-in way to build and share reports. A dashboard has a Data tab, where you add datasets as SQL queries or Unity Catalog tables and metric views, and canvas pages where you add visualizations, text and filters. Visuals are built by picking fields, or by describing the chart in plain language and letting the AI assistant draft it. When ready, you Publish the dashboard, choosing whether viewers see data with your embedded credentials or their own permissions, share it with users or the whole account, and set a schedule to refresh it and email subscribers.
Example: Meera builds Sales Overview with a dataset on main.sales.orders joined to customers, a bar chart of sales by city, a counter for total sales and a date range filter, then publishes it to the sales team with a daily 7 AM refresh.
Why we use it
Teams need governed, fast reports right next to the data without exporting to another tool. AI/BI dashboards run on SQL warehouses, respect Unity Catalog permissions and can be shared with people who do not have a Databricks login through account-level sharing.
When to use it
Use AI/BI dashboards for operational and team reporting on lakehouse data. Use Power BI or Tableau when the company already standardises on them or needs complex semantic models. Use Genie Agents for ad-hoc natural-language questions instead of more dashboard pages.
How it works
- + New → Dashboard
- Data tab: add a SQL dataset or pick a Unity Catalog table
- Canvas: Add a visualization, choose type and fields
- Add filters, a counter and text for context
- Should viewers use your data access? Yes: Publish with embedded credentials. No: Publish without, so viewers need SELECT.
- Share and add a refresh schedule
Syntax
UI: + New → Dashboard (or Dashboards → Create dashboard)
Data tab → Add SQL dataset / Add data source → canvas → Add a visualization → pick fields
Publish (embed credentials or not) → Share → Schedule
As code: dashboards in Declarative Automation Bundles (.lvdash.json)Example
-- Dataset 'orders_by_city' on the Data tab
SELECT c.city, c.segment, o.order_date, o.amount
FROM main.sales.orders o
JOIN main.sales.customers c ON o.customer_id = c.customer_id
WHERE o.order_date >= :start_date;
-- Canvas widgets built on that dataset:
-- Counter: SUM(amount) -> Total sales
-- Bar: X = city, Y = SUM(amount) -> Sales by city
-- Filter: date range on order_date, single-select on segmentKey points
- Former names: Lakeview dashboards; legacy DBSQL dashboards are retired
- Datasets live on the Data tab; one dataset can feed many visuals
- Publish creates the viewer version; your draft stays editable
- Embedded credentials let viewers see data through the publisher's access
- Schedules refresh the dashboard and can email a snapshot to subscribers
- Filters and parameters let viewers slice data without editing SQL