What’s New in AI/BI - July 2025 Roundup

Introduction

Since announcing the General Availability of AI/BI Genie and introducing Databricks One at our most recent Data + AI Summit, we’ve continued to improve how organizations analyze and act on data. This month’s roundup highlights new capabilities in AI/BI Dashboards and Genie — from deeper semantic tooling to easier customization and better user experience.

These updates are part of our mission to make business intelligence more accurate, flexible, and accessible. For context, Databricks AI/BI is a built-in suite of BI capabilities in Databricks SQL that lets users ask questions in natural language, build dashboards, and share insights — all powered by the Databricks Data Intelligence Platform.

What’s New?

As usual, let’s start by looking at some of the major new headlines for both AI/BI Dashboards and Genie.

Dashboard Themes

Theme customization—one of the most frequently requested features from customers—is now available in AI/BI Dashboards. Authors can now configure color palettes and font styles to better align dashboards with their organization’s branding standards. These settings can be adjusted from the dashboard’s settings panel, where users can choose a preset theme or define a custom one—including colors for text, background, widgets, and chart series, as well as title alignment and font styles. Theme changes apply consistently across the entire dashboard and are preserved during export and import, enabling scalable, brand-aligned reporting across teams.

Global Dashboard Filters

Dashboard authors can now define Global Filters that apply across all pages of a dashboard. Accessed by clicking the filter icon in the dashboard toolbar, Global Filters open a configuration panel where you can add interactive filter widgets scoped across the entire dashboard. Choose from field-based filters (e.g., date pickers, categorical dropdowns) or parameter-driven filters, and specify default values to control behavior on initial dashboard loads. This feature dramatically simplifies multi-page dashboard design—enabling consistent, intuitive navigation and reducing redundant setup—while giving users a centralized filter experience.

Dashboard Drill Through

Drill Through offers users a powerful new way to explore dashboards by applying selections from one page as filters on another. With this feature, viewers can right-click on a data point in a source visualization—such as a bar, pie chart and navigate to a target page where their selection is applied as a local filter. For example, selecting “United States” in a sales-by-country bar chart on a summary page can automatically filter the detail page to show only U.S. data. Drill Through unlocks new workflows for investigative analysis, helping users move seamlessly from high-level KPIs to deeper diagnostic insights without needing to reset filters or manually repeat selections.

Dashboards File Upload

You can now upload Excel, CSV, or TSV files directly into Unity Catalog from the Dashboards data tab. To define a dataset from a file, after selecting a dashboard draft, navigate to Data, choose Upload file, and specify the target Catalog, Schema, and Table name. You choose whether to create a new table or overwrite an existing one—just ensure you have MANAGE privileges on the target schema. Once uploaded, the file becomes a first-class dataset: query it using SQL, visualize it immediately, or combine it with other Unity Catalog data—all without leaving the dashboard interface. This empowers users to explore ad hoc or external data seamlessly, reducing friction in the analytics workflow.

Forecasting for Dashboard Line Charts

Now available in Public Preview, AI/BI Dashboards now leverage the ai_forecast function to let users generate forecasts directly from line charts, transforming historical trends data into forward-looking insights. To use it, select any line chart with a temporal x‑axis and a numeric y‑axis, then click the “+ Forecast” button in the visualization editor and choose “Add AI Forecast”. This creates a new chart that seamlessly overlays a forecasted trend line. The underlying ai_forecast function calculates projections based on your selected data range and aggregation logic. This feature makes it easy to anticipate future patterns—whether that’s sales performance, web traffic, or operational metrics—and incorporate proactive insight into your dashboards.

Enhanced Genie Knowledge Store

AI/BI Genie’s knowledge store is a foundational advancement for improving the quality, relevance, and explainability of answers returned by Genie. Available within each Genie space, the knowledge store gives analysts a set of tools to fine-tune how Genie understands data—enabling more trustworthy natural language interactions without having to modify upstream Unity Catalog assets to start curating the space.

Popular queries are now automatically surfaced when new tables are added, giving business users suggested Example SQL based on frequently run queries that the author has access to.

Authors can also curate and edit dataset metadata for their Genie space with several new capabilities:

• Local table and column descriptions allow authors to quickly iterate on table metadata within a Genie space, without needing to overwrite the descriptions in Unity Catalog. This is particularly helpful when an author lacks permission to modify Unity Catalog assets, wants to test different descriptions before making a modification, or wants descriptions that improve Genie accuracy but don’t belong as a central definition.

• Column synonyms make it easier for Genie to map varied business terms to the correct technical fields.

• Column show/hide lets authors declutter the semantic surface by excluding duplicate, irrelevant or overly technical fields, without requiring new tables or views.

• Example values help Genie better understand how your data is formatted across all different column types (strings, numerics, dates, etc).
• Value dictionaries help Genie map user prompts to unique string values in your data (e.g. mapping a user prompt for “UK” to “country = United Kingdom”). Genie space authors can now add value dictionaries for up to 60 columns and 1,024 unique values per column.

• Join relationships can now be declared for any dataset to specify joinable datasets, SQL JOIN expressions, and join types to be used. This helps authors who lack permissions to set Primary/Foreign Keys in Unity Catalog or are working with views that don’t support Primary/Foreign Keys. Existing Primary/Foreign Keys from Unity Catalog will automatically sync to reduce risk of duplication/conflict.

Together, these new features give authors significantly more control over Genie’s behavior and enable more accurate, context-aware responses—marking an important step toward delivering governed, reliable, and business-aligned conversational analytics. These local versions of table metadata let space authors quickly iterate and define context useful for Genie without needing to update Unity Catalog globally. For more information on Genie’s knowledge store, please check out the product documentation.

Other notable features

We are constantly working to expand and improve AI/BI Dashboards and Genie with new features and capabilities. Our ongoing efforts highlight our dedication to making AI/BI more user-friendly, robust, and impactful in supporting your organization's data-driven decision-making. Some noteworthy updates are included below; for a complete list of features and fixes, please refer to the AI/BI product release notes.

AI/BI Dashboards

• Choropleth Maps. Choropleth maps allow data to be visualized geographically by coloring regions based on aggregated values. Ideal for showing patterns like revenue by state or incidents by country, this new map type supports U.S. and international geographies and adds a spatial layer to decision-making in dashboards.

• Parameterized Dashboard Schedules. Authors can now define named dashboard schedules with custom filter parameters, enabling targeted report delivery without duplicating dashboards. This is ideal for teams distributing reports to different regions, business units, or customer segments — all while maintaining a single source of truth and reducing maintenance overhead. For more information on creating and maintaining schedules, please refer to the product documentation.
• Schedule Without Embedded Credentials (Run as Viewer). Scheduled dashboard deliveries can now be executed using the recipient’s viewer permissions instead of embedding the author’s credentials. This enhances security and access governance, ensuring that each user sees only the data they’re entitled to—and aligns delivery behavior with enterprise compliance policies.
• Define Custom Calculations Over a Range: AGGREGATE OVER measures are now supported in custom calculations, empowering analysts to create rankings, running totals, or moving averages. This removes the need to define window functions within the SQL query directly, enabling filtering and aggregations on the same base dataset across multiple widgets.

AI/BI Genie

• Editable Results Filters. Genie users can now refine Genie answers by editing filters directly within the results panel. For example, after receiving an answer filtered by a specific region or product, users can quickly adjust the filter values and regenerate results — helping users quickly correct queries and accelerating time to insight.

• Point Maps Now Supported. Genie now supports point maps in visual responses, enabling users to view geographic data plotted as individual points based on coordinates. This enhancement brings location intelligence to conversational analytics and helps users visualize patterns like store locations, service requests, or delivery routes.

• Follow-up Questions for Text Responses. Genie now suggests follow-up questions for text-based responses that don’t include generated SQL or result sets. Previously, follow-up prompts appeared only after responses with data output. Now, even responses like “Explain the dataset” include suggested follow-up questions to help users discover new questions they can ask and deepen their exploration within a Genie space.

• Genie API List and Delete Operations. New API endpoints let developers programmatically list and delete Genie spaces and conversations within a space— helping organizations automate governance, manage content at scale, and integrate Genie into existing workflows and tooling. These APIs are an important foundation for teams rolling out Genie across large enterprises.

What’s Next?

We’re continuing to build toward a more unified, intelligent, and accessible analytics experience across Databricks AI/BI. Here’s a quick look at what’s coming:

• Business semantics across Genie and Dashboards — Already in Public Preview for Dashboards, Metric Views will soon power Genie as well, bringing trusted KPIs into natural language workflows and closing the loop between visual and conversational analytics.
• APIs for embedded Genie experiences — New APIs will let you export/import spaces and collect answer feedback, unlocking scalable deployment and tighter integration into your own apps.
• Smarter Dashboards and easier sharing — Soon you’ll see a redesigned counter visualization, a new Apply button for filters, and support for CSV attachments in email subscriptions — all designed to boost usability and reach across your organization.

We’ll be diving deeper into AI/BI semantics in upcoming blogs and resources — so stay tuned!

Learn More

If you’re looking to experience the latest AI/BI capabilities firsthand, the options below are designed to guide you through.

• Free Trial: Get hands-on experience by signing up for a free trial.
• Documentation: Dive deeper into the details with our documentation.
• Webpage: Visit our webpage to learn more.
• Demos: Watch our demo videos, take product tours and get hands-on tutorials to see these AI/BI in action.
• Training: Get started with free product training through Databricks Academy
• eBook: Download the Business Intelligence meets AI eBook