Session

Getting the Most out of AI in BI: From Dashboard Creation to Intelligent Insights

Overview

Experience	In Person
Track	Analytics & BI
Industry	Enterprise Technology
Technologies	Genie
Skill Level	Intermediate

AI is rapidly reshaping business intelligence but for many teams, it’s still unclear how to move beyond isolated features to real, end-to-end impact.

In Databricks, AI is embedded across the entire BI workflow. From using Genie Code to build and maintain dashboards, to applying AI-driven capabilities like forecasting and key driver analysis, to enabling conversational exploration directly within dashboards, AI becomes part of how analytics is created, delivered, and used.

This shift allows teams to move faster, reduce manual work, and unlock deeper insights without adding complexity.

In this session, we’ll walk through how to apply AI across the full BI lifecycle:

- How to use Genie Code to accelerate dashboard creation and maintenance

- How to leverage built-in AI features such as forecasting and intelligent insights

- How to embed Genie directly into dashboards for conversational exploration

- Best practices for combining AI and BI to drive adoption and real business impact

Session Speakers

IMAGE COMING SOON

Alex Lichen

/Sr. Product Manager
Databricks

IMAGE COMING SOON

April Dai

/Engineering Manager
Databricks

Full Summary

Beyond dashboards: how AI/BI Dashboards and Genie reshape self-service analytics

BI teams face a two-sided squeeze: a backlog of bespoke dashboard asks and an emerging surge of ungoverned, AI-spun artifacts. The session proposes a governed model where core teams invest in trusted semantics and a handful of high-value dashboards, while AI handles the long tail of questions through a unified, controlled experience.

The approach builds on the business intelligence and analytics approach for the AI era Databricks has been shaping. Teams evaluating how AI is changing the analyst role will find grounding in how AI is transforming data analytics across modern data organizations.

The idea is simple and powerful. Put definitions, lineage, and permissions at the center, then let AI operate on top of the same governed context so answers stay accurate, transparent, and repeatable.

Shift the BI stack to a unified semantic foundation

Traditional BI separates where agents run from where metadata and governance live. On the Databricks Platform, that split closes. Unity Catalog holds the semantic layer, and the Genie Ontology builds a context graph that identifies authoritative definitions for core business concepts such as revenue and ARR. Answers remain grounded in governed meaning instead of ad hoc SQL over raw tables.

BI teams still publish key dashboards, but the long tail moves to self-service AI. Users ask follow-ups or net-new questions without spawning more one-off reports, and every answer inherits the same lineage, definitions, and access controls.

Authoring with Genie Code, from intent to governed assets

Genie Code turns written requirements and reference screenshots into working dashboards. It selects datasets, generates Unity Catalog metric views with measures, dimensions, and relationships, and can materialize common grains for speed. Local, dashboard-scoped metric views let authors model quickly without upfront catalog changes, while YAML and UI views keep definitions transparent.

Verification and automatic error recovery reduce cleanup after generation. When built-in visuals are insufficient, Genie Code produces custom charts via VegaLite specifications. Skills, including an Import BI skill for Tableau and Power BI, help teams bring existing assets into the governed environment for enhancement rather than rebuilds.

Consumption with embedded Genie and Genie One

Published AI/BI Dashboards become interactive analysis surfaces. Viewers can ask why a metric moved, see the drivers, and review the reasoning steps and SQL. Answers reflect current filters and visual context and respect dashboard permissions.

Genie One gives business users a single place to ask questions without hunting for the right asset. It searches across entitled AI/BI Dashboards, agents, and tables, assembles editable documents, and supports scheduled delivery by email, Slack, or Teams. Reusable skills capture prompt templates and calculation steps, while Genie Agents provide curated, narrow context with custom instructions for repeatable expert answers. MCP support brings in relevant content from systems such as Google Drive and SharePoint when needed.

Roadmap focus for AI/BI Dashboards and Genie

Speakers flagged several forward-looking themes that may evolve based on feedback.

AI/BI Dashboards
- - AI-elevated authoring and automatic insight surfacing
- - Richer visuals and interactivity, including drill downs
- - Enterprise reporting, such as version history and expanded delivery destinations
- - Performance tied to Raiden and metric view materialization
- - AI-assisted dataset deduplication to reconcile conflicting definitions in Unity Catalog
Genie One
- - Deeper context via indexing, ontology, MCP, and Lakeflow Connect
- - More outcomes, including documents and agents that can act on external systems
- - Stronger governance and admin visibility, plus branding controls
- - Mobile experience improvements, including push notifications and photo or file upload

Governed semantics are the lever for scalable self-service

Generative analytics scales when AI inherits the same semantics, governance, and lineage as dashboards and data models. Unify context first, then let AI serve the long tail. The result is faster answers, fewer brittle assets, and a single standard for trust.

FAQ

Genie Ontology is a context graph that learns authoritative definitions for core business concepts across the Databricks Platform and connected systems. It gives AI grounded, consistent meaning, so answers reflect trusted metrics and lineage rather than guesses over raw tables.

Genie Code helps technical authors generate dashboards, metric views, and data models. Genie One is the business user entry point for asking questions across entitled assets and receiving editable documents. Genie Agents are curated, narrow contexts with specific data and instructions for repeatable expert answers.

Yes. Genie only considers AI/BI Dashboards, tables, and agents the user is authorized to access. Interactions on dashboards inherit dashboard permissions, and the product exposes reasoning steps and SQL for transparency.

Yes. An Import BI skill in Genie Code brings Tableau or Power BI dashboards into the Databricks Platform so teams can continue enhancing them with AI-powered capabilities without starting from scratch.

The session mixed available features with future themes. Branding controls, scheduled deliveries, MCP support, and the mobile app are available today. Roadmap items include shareable skills, expanded delivery destinations, richer custom layouts, AI-assisted dataset deduplication, and mobile push notifications. Many items are subject to change based on customer feedback.