Databricks Genie Major Update: Context, Data Ingestion, and Automation Get a Full Upgrade

Databricks' AI assistant Genie has received a broad set of feature upgrades spanning four dimensions: contextual understanding, data ingestion, collaboration, and automation. This update moves Genie from a simple Q&A tool toward an intelligent assistant that fits more naturally into real enterprise workflows. Here's a breakdown of the core changes and what they mean for data teams.

Business-Aware Context: Genie Ontology Enabled by Default

A common pain point when using AI to query data has been the model's lack of understanding of enterprise business context — it knows SQL syntax, but doesn't know what "active users" or "net revenue" specifically means within your company. In this update, Genie Ontology is enabled by default, meaning every query automatically carries business-aware context.

The underlying logic is to bridge the enterprise semantic layer with AI querying capabilities. When a model understands the business definitions behind the data, the query results it generates are far more likely to reflect the user's actual intent — rather than mechanically matching field names. For non-technical business users, this lowers the bar for "asking the right question."

Ontology in the data world refers to a formal description of business concepts and their relationships — it defines what an "active user" is, how "net revenue" is calculated, which dimensions share the same reporting definition, and so on. Think of it as a company-specific semantic dictionary. Traditionally, this knowledge is scattered across data dictionaries, BI-layer metric definitions, or employees' tacit expertise. Without this layer, large language models can only infer meaning from field names, which often leads to semantic drift. Genie Ontology works by structurally injecting an enterprise's existing semantic layer — typically sourced from table descriptions in Unity Catalog, metric definitions, or dbt model annotations — into every AI inference step. This allows the model to align with the company's own business logic when generating SQL or interpreting results, rather than relying on generic natural language guesswork.

Workspace Instructions: Applying Data Standards to Every Query

Another noteworthy capability is workspace instructions. Organizations can configure their data conventions — such as naming rules, metric definitions, and filtering logic — as unified instructions that are automatically applied to every prompt.

This addresses the consistency problem. In large organizations, different people may interpret the same metric differently, and AI-generated results can easily diverge. By enforcing workspace-level instructions as a universal constraint, query outputs across a team can be held to the same standard, reducing alignment overhead downstream.

Expanded Data Ingestion: Multi-Format Files and Native Unity Catalog Access

This update significantly broadens the data sources Genie can work with. Users can now upload Word documents, images, CSVs, spreadsheets, and PDFs directly into their workflow, enabling AI-powered analysis on unstructured or semi-structured data. This extends Genie's use cases from database querying into everyday document processing.

More critically, there is now native support for Unity Catalog tables. Users can query Unity Catalog tables directly within Genie, use schema previews to quickly understand table structure, and request access permissions with a single click. This connects data governance with data querying — preserving the security boundaries of access control while eliminating the friction of constantly switching between tools.

Unity Catalog is Databricks' unified data and AI governance layer, covering assets such as tables, files, ML models, and functions. It provides fine-grained permission controls, data lineage tracking, and audit logs across workspaces. Before Unity Catalog, permission management in Databricks was siloed per workspace, and cross-team data sharing required tedious manual synchronization. Unity Catalog centralizes governance logic through a unified Metastore, allowing administrators to define "who can access which columns of which table" in one place and automatically propagate those policies to all connected workspaces. Genie's native integration with Unity Catalog means users no longer need to leave the conversation interface to request permissions or look up documentation elsewhere. The entire workflow — discover data → understand structure → request access → start querying — can now be completed within a single context, dramatically reducing the friction cost of data exploration.

Automation: Scheduled Tasks with References to Historical Runs

On the automation front, Genie now supports scheduled tasks, enabling repetitive work to be configured as periodic, automatically executed jobs. A particularly interesting design choice is that these tasks can reference past run records, ensuring consistency in output results.

This is especially useful for teams that need to generate reports or monitor metrics on a regular cadence. Referencing historical runs not only eliminates the need to reconfigure each time, but also keeps output formats and metric definitions stable across multiple runs — reducing the burden of manual review.

The Product Philosophy Behind the Update

Taken together, these updates reveal a clear shift in how Databricks is positioning Genie: from "an AI that can answer data questions" to "an AI collaborator embedded in enterprise data workflows."

Context and workspace instructions address accuracy of understanding. Multi-format ingestion and native Unity Catalog access address data accessibility. Scheduled tasks address efficiency and repeatability. These three capability clusters converge on a single goal — making AI a genuine part of a data team's daily operations, not just a flashy demo tool.

According to a demonstration by Databricks team member Youssef Mrini, the full feature set can be walked through in under three minutes — a signal that the design prioritizes an out-of-the-box experience over complex configuration requirements.

What This Means for Data Teams

For organizations already using the Databricks platform, this update lowers the technical barrier for business users to self-serve data queries, while maintaining data definition consistency through ontology and workspace instructions. For data engineers, deep integration with Unity Catalog means governance and analysis can be completed within the same interface, with a smoother permission request process.

Overall, the value of this upgrade lies not in any single standout feature, but in how it integrates previously scattered capabilities into a coherent workflow. As competition among enterprise AI assistants intensifies, the platforms that best embed AI into existing data stacks and governance frameworks are the ones most likely to earn lasting adoption from data teams.