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

As autumn arrives, Databricks AI/BI continues to advance—bringing faster, smarter insights to everyone. Over the past few months, both AI/BI Dashboards and Genie have gained new capabilities to make your analytics more intuitive, governed, and intelligent. Highlights include the ability to embed dashboards in external applications, apply tags and certification for stronger governance, and explore data through Genie’s expanded APIs and improved reasoning experience.

Built directly into Databricks, AI/BI unifies visualization, natural language exploration, and governance in one platform—so users can analyze and share insights without leaving the Data Intelligence Platform. Whether you’re building dashboards, asking Genie to uncover the “why” behind trends, or scaling insights across teams, AI/BI turns data into decisions with fewer tools, less friction, and a lot more intelligence.

What’s New?

To start, here’s a look at the standout updates that make this release one of our most feature-packed yet, covering both AI/BI Dashboards and Genie.

Dashboards embedding for external users (Public Preview)

With embedding for external users, AI/BI Dashboards can now be securely embedded in applications built for audiences outside your organization, such as customers, partners, or vendors, without requiring Databricks accounts. These embedded dashboards keep data governed through Unity Catalog, and access is managed using a service principal that authenticates your application via OAuth tokens, giving developers precise control over data access and permissions. 

Each viewer’s experience can even be personalized using user-scoped tokens that dynamically filter results based on their profile or role. And because AI/BI uses the same consumption-based pricing model for embedded and internal dashboards, you only pay for the underlying compute—no additional viewer or per-view fees for embedded analytics. Whether it’s a customer portal showing product usage or a partner dashboard highlighting key performance metrics, embedding for external users makes it simple to extend trusted insights beyond your Databricks environment.

Please review our product documentation to learn more about how to set up dashboard embedding for external users.

Dashboards Top/Bottom N Bar Chart Categories

Bar charts can now highlight just the best (or worst) performing categories in your data. Instead of displaying every category on the axis, you can choose to show only the top or bottom N categories based on the metric you’re plotting. To try it, open the kebab menu for the bar chart’s categorical axis dimension and set the Default number of categories to the desired value. This feature provides an ideal way to focus on outliers, like your top ten customers or your bottom five products, without having to manually filter your dataset.

Embedded credentials for Dashboard Genie

Genie spaces that are automatically generated from dashboards now support embedded credentials, ensuring that all Genie queries run using the same data permissions as the dashboard itself. When a dashboard is published with embedded credentials, Genie uses the publisher’s credentials to execute queries, allowing viewers to explore related questions without needing their own data access. Dashboards published without embedded credentials continue to run queries using the viewer’s permissions, preserving your organization’s governance model. This enhancement also brings Genie’s conversational capabilities to externally embedded dashboards, enabling users to securely ask follow-on questions and get AI-powered insights directly from within the applications they already work in — without needing to switch to Databricks.

Subscribe Slack Channels to Dashboard Snapshots 

AI/BI dashboards can now deliver scheduled snapshots directly to Slack channels, bringing insights into the tools your teams use every day. Each post includes a PNG image preview of the dashboard, a direct link to open it in Databricks, and a PDF snapshot attachment in the message thread. Slack delivery works alongside your existing dashboard schedules—after each scheduled refresh completes, the latest snapshot is automatically shared with your selected channels. Just like email subscriptions, it ensures stakeholders get timely updates and trusted insights pushed straight to them, without needing to always check their dashboards manually.

170+ new functions added in custom calculations

Over 170 new functions have been recently added to custom calculations, including support for AGGREGATE OVER to compute moving aggregations and FILTER(WHERE …) to apply conditional filters without complex expressions. Custom calculations also now recognize predicates such as IN, BETWEEN, and pattern matching (LIKE, ILIKE, RLIKE), giving authors more flexibility in shaping data logic. Check out our product documentation to see a full list of supported functions in custom calculations.

Genie Conversation API enhancements

Genie’s API capabilities have expanded significantly, making it easier to manage, monitor, and integrate conversational analytics into your own applications. New endpoints now allow you to record user feedback (thumbs up or down), retrieve suggested follow-up questions, delete messages, list all conversations and messages for review, and return additional metadata, such as the associated warehouse ID, in API responses. Additionally, Genie has adopted the Databricks Permission APIs for consistent permission handling across the platform. Together, these enhancements simplify how teams programmatically refine Genie’s behavior, gather feedback, and embed its conversational intelligence wherever it’s needed. To learn more, see the documentation on using the Genie API and setting up a Genie space.

Genie benchmarking improvements

Genie’s benchmarking tools have been refreshed to make it simpler to evaluate your space’s accuracy and apply a higher degree of rigor. It’s now much simpler to create benchmark sets: you can create a benchmark directly from chat responses by choosing Add as benchmark from the actions menu and from other user prompts via the Monitoring page. The Add Benchmark modal also has a Generate SQL option that writes a draft answer for you.  We have also improved the scoring function so that it more accurately judges accuracy and explains incorrect answers. Additionally, benchmarks now run asynchronously, so you can work while they complete.

To help space authors understand when to run benchmarks, Genie also now prompts you to re-run benchmarks when it detects a substantive change in Genie context. Once a benchmark run is complete, you can save a better Genie-generated answer as the new ground truth.

Enhanced Genie Knowledge Store

The Genie Knowledge Store now allows authors to define structured business semantics to enhance Genie’s accuracy. To improve Genie’s understanding of table relationships, we’ve improved the ability for space authors to define local join relationships inside a Genie Knowledge Store by selecting tables, choosing the join columns and declaring the relationship type (many to one, one to many or one to one). Genie automatically adds table aliases when there are multiple joins or self-joins. 

We’ve also released SQL expressions, a guided way to directly teach Genie about common filters, measures, and dimensions. 

Finally, Genie now includes Knowledge Extraction, a new feature that helps it learn directly from user feedback. When a user gives a thumbs-up to a generated query, Genie analyzes that interaction to propose knowledge snippets—like potential measures, dimensions, or filters. Space authors can review and approve these snippets before adding them to the Knowledge Store, helping Genie continuously refine its understanding of your data through real-world use.

Collectively, the above Knowledge Store updates give you much more control over how Genie understands and uses your data.

Genie user experience improvements

We’ve recently added some key updates to make interacting with Genie feel far more intuitive.  First, Genie exposes its reasoning with a new thinking steps panel that shows how it interpreted your prompt and which tables and SQL queries it used. This helps non-technical users review the accuracy of answers without needing to understand SQL. Second, we now return natural language summaries of the generated query results to help users quickly understand their data insights. Together, these changes bring more clarity to every conversation. 

You’ll also notice a few quality‑of‑life improvements, such as feedback options that persist across follow‑up questions and the ability to edit parameters for dates and numbers. Overall, these updates make it easier than ever to interpret, understand and refine Genie’s answers.

Tags and certification for Dashboards and Genie spaces.

You can now apply governed tags and system certification labels to both AI/BI Dashboards and Genie spaces, making it easier to organize, classify,  and discover your trusted analytics assets. Governed tags ensure consistent labeling across your workspace through centrally defined tag policies in Unity Catalog — helping teams categorize dashboards and spaces by concepts like department, project, or business domain. Certification, a system-governed tag, allows data stewards to flag assets as certified or deprecated, signaling reliability and lifecycle status at a glance. Together, Dashboard tags and Genie space tags bring powerful governance and discoverability to AI/BI.

The image above shows a Dashboard with applied tags and certification, while the image below illustrates the same for a Genie.

Other notable features

Beyond the major features, AI/BI Dashboards and Genie continue to receive a steady stream of enhancements that improve reliability, performance, and usability. Below are several notable updates from recent months, with full details available in the AI/BI release notes.

AI/BI Dashboards

• Active filter bar: A new filter bar now appears at the top of dashboards, showing all active filters applied, whether global or page‑level, including any cross‑filters. This creates a simple way for viewers to see which filters are applied for their current view at a glance.

• Undo and redo support: Editing dashboards just got much safer. Authors can now undo or redo canvas actions using keyboard shortcuts or the toolbar (Cmd/Ctr+Z for undo, Cmd/Ctr + Shift + Z OR Cmd/Ctr + Y to redo), making it easy to revert accidental changes or repeat a layout tweak.
• Improved text widget editor: The text widget now features an upgraded editor that makes it easier to format text and include links or images. Authors can add headings, visuals, and hyperlinks directly within the widget to create more polished, on-brand content and connect readers to related information.
• Funnel chart support: A new funnel chart visualization lets you display conversion flows and sequential drop-offs directly within AI/BI Dashboards. This makes it easy to visualize how users or events progress through each stage of a process, from beginning to end.

• Waterfall chart support. Waterfall charts provide a clear way to visualize how sequential positive and negative changes contribute to a total. This is useful for illustrating trends like revenue breakdowns or expense changes, helping viewers quickly understand how each factor drives the overall result.

• Copy/paste pages and widgets between dashboards: You can now copy and paste entire pages or individual widgets between dashboards. Simply select a page or widget, and use keyboard shortcuts (Cmd/Ctrl+C and Cmd/Ctrl+Z) to paste assets to another dashboard. This makes it easy to reuse layouts, reports or filters without recreating them from scratch.
• Query insights: Query profiles are now available in the dataset editor to help you analyze query performance and optimize your dashboards with confidence.
• Simplified publish dialog: The publish dialog has been redesigned and decluttered. The Genie integration toggle has also moved into dashboard settings, making dashboard publishing clearer.

AI/BI Genie

• Metric Views + Tables in the Same Genie Space: Users can now analyze metric views and tables together within a single Genie space, enabling seamless exploration of both aggregated and granular data without switching contexts.
• Improved Metric Precision: Genie’s SQL generation has been refined to return cleaner, more accurate percentage results (e.g., conversion rates), consistently rounded to two digits for readability.
• Improved Context Handling: We’ve made a number of improvements behind the scenes to improve Genie’s ability to manage the context from instructions, previous user prompts, and table metadata to accurately determine user intent behind their prompts.

What’s Next for AI/BI?

AI/BI continues to evolve through ongoing feedback and innovation. In the months ahead, we’ll focus on advancing both Dashboards and Genie to make analytics even simpler, smarter, and more accessible for everyone. Here are a few key areas of investment on the roadmap:

• Genie Research Agent Mode. A new Research Agent Mode is planned to help Genie tackle complex “why” and “how” questions by using multi-step reasoning, testing multiple hypotheses in parallel, and returning conclusions with citations.
• Genie - Revamped clarification questions for ambiguous prompts. Genie will soon handle ambiguous prompts more intelligently by recognizing when a user’s question could have multiple interpretations and proactively asking clarifying questions. This enhancement makes conversations feel more natural while ensuring that Genie’s answers are contextually accurate and relevant to the user’s true intent.
• Dashboard Agent for Natural Language Authoring. This will introduce a conversational authoring experience, allowing users to create and edit entire dashboards and widgets with natural language. Simply, by describing what they want to see, authors will be able to add charts, filters, and calculations—no SQL required.
• Low-code authoring for Metric Views. A new low-code interface will soon allow for the creation and editing of Metric Views. This experience will make it easier for analysts and business users to define common measures, dimensions, and filters without having to touch a YAML file.
• More Dashboards interactivity and presentation features. We’re also investing in how users interact with and present insights in their dashboards. Upcoming enhancements include greater control over filters with bulk apply actions, per-widget and relative date options, and custom range selections. We’re also introducing layout and design improvements like a denser canvas grid, workspace-level themes, and new drill-down navigation.

Stay tuned for more updates as we continue to expand what’s possible with Databricks AI/BI.

Learn more

If you are ready to explore the latest in AI/BI, you can choose any of the following options:

• Free Edition and Trial: Get hands-on experience by signing up for our free edition or 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

And keep an eye out - we have more innovations coming soon!