“The lakehouse is no longer just an analytical foundation. It is becoming the decision backbone that feeds and powers enterprise AI.”, as HyperFRAME Research notes. 

Traditional analytics often copied data from operational systems into separate analytic platforms. That approach can add latency and cost while fragmenting security controls. AI agents need fast access to live enterprise data across both Oracle and non-Oracle databases, Apache Iceberg tables, object storage, external catalogs, and applications.  

But access alone is not enough. To make informed decisions and take trusted action, agents need to understand the business meaning of the data, how it relates across sources, and the organization’s rules and policies. Without that context, answers may sound right but still be wrong for the business. 

Oracle Autonomous AI Lakehouse brings AI, business context, and unified governance to live data across Oracle and non-Oracle data sources, including third-party platforms such as Snowflake and Databricks. Its open architecture integrates with customers’ existing data platforms and tools, enabling users and AI agents to access and analyze data in place, reducing the need to copy or move data into separate environments. Your data can stay where it is. AI comes to it.  

Autonomous AI Lakehouse is available in Oracle Cloud Infrastructure (OCI), Amazon Web Services (AWS), Microsoft Azure, Google Cloud, Oracle Exadata Cloud@Customer, and OCI Dedicated Region.  

New and expanded capabilities

Oracle today announced new and expanded Autonomous AI Lakehouse capabilities that help organizations bring business context to users and AI agents, perform multidimensional simulations, and strengthen governance across the enterprise data estate. They span three key areas: 

  • Business-aware Agentic AI: enabling agents to understand live enterprise data in its business context 
  • Lakehouse 360: providing an end-to-end, unified lakehouse experience for analytics, AI, data engineering, and business simulations 
  • Trusted, governed AI: helping organizations manage data access through role-based controls and lineage information across data sources 

Let’s review the new and expanded capabilities within each key area in more detail. 

Business-aware Agentic AI:

AI Data Catalog  

AI Data Catalog provides a centralized view of enterprise data by allowing customers to mount Oracle and non-Oracle databases, external catalogs, Iceberg tables, and data in object store without duplicating data. Customers can connect to external catalogs such as AWS Glue, Databricks Unity Catalog, and Snowflake Horizon Catalog, bringing discovery, business semantics, lineage, and governance together across distributed sources. These data sources can be queried live via the AI Data Catalog using the full power of the Oracle engine. 

The AI Data Catalog continuously enriches metadata with lineage, workload patterns, relationships, sensitive-data context, business semantics and retrieval signals. Agentic search and agentic pruning help users and AI agents navigate large, complex environments by identifying the most relevant assets for a question or workflow. Hierarchical pruning progressively narrows the search from catalogs and schemas to relevant tables and columns, using business meaning and relationships to identify the context needed for each question. This focused context helps improve the quality and speed of discovery, natural-language analytics, and agent workflows while helping to reduce costs by limiting the metadata agents must process. 

Business Semantic Layer  

Autonomous AI Lakehouse’s Business Semantic Layer enables customers to add and enrich business meaning across all connected data sources, providing a shared semantic foundation for users, applications, and AI agents. The semantic layer lets users create annotations, import existing Oracle Database SQL annotations, and generate AI-enriched table and column descriptions. Shared business definitions and semantic context help users and AI agents interpret data consistently across connected sources, analytical tools, and agent workflows. 

Customers can also define complex multidimensional and hierarchical data models using AI-suggested Oracle Analytic Views. These views capture dimensions, measures, metrics, hierarchies, calculations, and KPIs. By encapsulating joins, aggregations, hierarchy logic, and business calculations inside reusable Analytic Views, the Business Semantic Layer helps reduce SQL complexity.  

Support for Apache Ossie enables customers to import semantic models from compatible systems and reuse existing business definitions across lakehouse data. This helps reduce the effort of recreating semantic models and promotes consistent business meaning across analytics, BI, and AI tools. Together, these capabilities help improve the accuracy and consistency of Select AI, NL2SQL, AI retrieval, and analytic insights while preserving existing Oracle Autonomous AI Database semantic investments and extending them to broader lakehouse data.  

Expanded native support for Iceberg data  

Autonomous AI Lakehouse extends Oracle AI Database’s SQL, AI, analytics, and enterprise data-management capabilities to open Apache Iceberg data. Enhancements to native Apache Iceberg support improve performance and expand functionality for customers accessing Iceberg tables through external catalogs or Oracle AI Data Catalog. Beyond querying and analyzing Iceberg data, users and AI agents can now create, populate, update, and delete Iceberg tables using familiar Oracle SQL. A consistent analytics model across Oracle and Iceberg data helps turn Iceberg tables into an active foundation for AI-powered analytics, agentic workflows, and reusable lakehouse data products.  

Customers can work with Iceberg tables managed in their existing supported catalogs, retaining their choice of catalog, object storage, and Iceberg-compatible engines. Expanded support for Iceberg V2 and V3 broadens compatibility across the open lakehouse ecosystem. These enhancements complement existing capabilities such as Vectors on Ice, allowing customers to combine relational analytics and vector-based AI workloads on Iceberg data.  

Oracle’s query optimizer uses available statistics from Iceberg metadata and data files, including compact statistics stored in Iceberg Puffin files, to select more efficient query execution plans, helping avoid expensive full-table scans during query execution. 

Support for on-premises object stores, including Oracle ZFS and Hitachi, adds deployment flexibility for customers that need to retain data in their own environments. Customers can preserve their existing Iceberg ecosystem while applying Oracle Autonomous AI Lakehouse performance, reliability, and AI capabilities to the same data, avoiding unnecessary copies and migrations. Governance through AI Data Catalog helps control access, while optimized writes, transaction recovery, and time travel support reliable updates, investigation of changes, and recovery. 

Live AI Hub  

Live AI Hub connects Autonomous AI Lakehouse to enterprise data and the business context needed to interpret it. Users and AI agents can query supported Oracle and non-Oracle databases, applications, and object stores in place. Semantic context from AI Data Catalog, external catalogs, ontologies, documents, and other sources helps Select AI and AI agents interpret natural-language questions and generate business-aware SQL. These connections are centrally managed through AI Data Catalog, together with access policies and credentials. Oracle’s SQL engine processes and optimizes queries across connected sources, bringing Oracle’s AI and analytics capabilities to distributed enterprise data. Customers can extend analytics and agentic AI without first moving existing Oracle data into a separate AI platform, preserving their investments. 

Pre-built and custom AI Agents  

Customers can use prebuilt agents available through Oracle AI Database Private Agent Factory and assistants for natural language to SQL (NL2SQL), data science, knowledge retrieval, deep research, and structured data analysis. They can also rapidly build custom agents and workflows that access lakehouse data using supported access controls to help generate insights and support business workflows, without first moving sensitive data into a separate AI platform.  

Lakehouse 360: Analytics, AI, Data Engineering, and business simulations

Data Studio (datastudio.oracle.com):  

Data Studio now provides a unified, AI-assisted workspace across multiple Oracle databases and lakehouse data sources, accessible through an OCI account. Users can discover, access, prepare, analyze, and manage data from a single workspace. Through integrated catalog and Data Transform pipeline capabilities, teams can use AI-assisted search to explore cataloged enterprise data, connect to more than 100 data sources, manage database resources, and support workflows spanning analytics, AI, data engineering, operations, and business simulations without switching between disconnected tools.  

Teams can work with data using natural language, SQL, or Python. Data Studio’s built-in AI Assistant, powered by Select AI, helps users query data conversationally, describe integration goals, create and refine Data Flows and Workflows, monitor and debug jobs, inspect database and OCI state, and access guidance from product documentation. Together, these capabilities give analysts, data engineers, DBAs, data scientists, and business users a common AI-assisted workspace for working across lakehouse data, workflows, and operations. 

AI-assisted Data Engineering  

Enables customers to build AI-assisted, low-code data pipelines that transform operational and other lakehouse data into governed, reusable data products. New Oracle Fusion integration through Spectra Data Extract streamlines extraction of Fusion data for transformation and publication. From Data Studio, teams can use natural language with the AI Assistant to build Transform pipelines. Powered by Oracle AI Database capabilities, these pipelines can cleanse, join, enrich, and publish operational, application, and lakehouse data without exporting database data for separate processing.  

Data engineers can use the Medallion pattern, including Bronze, Silver, and Gold layers, to prepare data for analytics, AI, and business simulations. For large-scale lakehouse processing, they can also build Spark pipelines in Transform, which orchestrates execution through Oracle AI Data Platform Workbench Spark or Amazon EMR. 

Business Simulation Engine  

Autonomous AI Lakehouse introduces a business simulation engine built on Oracle Essbase, bringing proven multidimensional simulation, planning, forecasting, and what-if analysis capabilities into the lakehouse. Users can build simulation models using lakehouse data, model scenarios with dynamic hierarchy rollups, advanced calculations, and allocations, and write results back to Oracle AI Database and Apache Iceberg tables.  These capabilities help teams evaluate the potential business impact of changing assumptions across revenue, demand, pricing, capacity, cost, risk, and other drivers. 

Business users can ask natural-language questions, generate simulation reports, and perform familiar Excel-based analysis through Oracle Smart View. The engine manages complex hierarchies and multidimensional calculations, while AI can suggest relevant Essbase cubes based on schemas and lakehouse context, which can help streamline the creation of planning and simulation models.  

Data Lake Accelerator and Lake Cache 

Provides a highly elastic, cost-effective way to scale out query processing and accelerate queries against data in object storage. The latest enhancements improve performance by 3X in our TPC-DS benchmark on a 100 TB dataset. The maximum  Data Lake Accelerator cluster size increases from 32 to 100 nodes, giving customers more capacity as they query larger datasets with Autonomous AI Lakehouse. 

Lake Cache complements Data Lake Accelerator by reducing the need to repeatedly access data in object storage and bringing Exadata performance to queries on that data. It caches frequently accessed data on Exadata flash, allowing subsequent queries to use cached data instead of reading it remotely. 

In benchmark testing, Lake Cache delivered up to 8X faster remote data access. For TPC-DS queries on Iceberg tables, query execution time decreased from 96 seconds without Lake Cache to 17 seconds with Lake Cache. Together,  Data Lake Accelerator and Lake Cache combine elastic query-processing capacity with fast access to external data. 

Trusted, governed AI across all your data

Centralized Security and Governance  

Autonomous AI Lakehouse  helps organizations manage data access and governance processes across on-premises and cloud data sources through role-based access controls in the AI Data Catalog. Customers can apply object-level access policies for end-users and agents across distributed Oracle databases, selected non-Oracle sources, and Iceberg tables. For example, customers can authorize access to individual Iceberg tables rather than granting broad access to an entire object-storage bucket, giving users and AI agents access to the specific data objects required for their work. 

This is especially valuable for customers with many databases, thousands of tables, and complex user communities. AI Data Catalog provides a centralized governance layer for discovery and access across lakehouse data, helping reduce the need to manage access policies separately for each data source. Centralized connection and credential management helps reduce administrative effort, while AI Data Catalog policies control which data users and AI agents can access through the lakehouse. These policies work alongside—not in place of—the access controls enforced by the underlying data sources. 

Automated Data Lineage  

Gives end-users and agents end-to-end visibility into lakehouse data, including where it came from, how it changed, and where it is used. Lineage is captured from database activity, Data Transform pipelines, Spark jobs, and connected lakehouse workflows, showing how data moves and changes across applications, databases, data lakes, and external catalogs.  

Autonomous AI Lakehouse embraces open standards such as OpenLineage to extend lineage visibility across the broader data ecosystem. OpenLineage support helps capture and exchange lineage events across open source and proprietary solutions such as Spark, Airflow, dbt, Databricks, Snowflake, and BigQuery. The lineage graph provides transparency into complex transformations, including joins, filters, business rules, and custom logic, so users and agents can trace data provenance, analyze change impact, and investigate issues more efficiently. For AI use cases, this evidence can help end-users, governance teams, and auditors investigate data provenance, assess generated outputs, and better understand the data used in analytics, NL2SQL, and agentic workflows.  

Credential vending 

Autonomous AI Lakehouse provides query engines and AI agents with temporary, table-scoped credentials to read and write Iceberg tables, based on access policies defined in the AI Data Catalog. Customers can use supported engines such as Spark and Trino, as well as agentic workflows, to access authorized Iceberg tables stored in object storage and managed through the AI Data Catalog or supported external catalogs. This preserves open lakehouse interoperability while applying customer-defined access policies through the AI Data Catalog, reducing reliance on broadly scoped, long-lived credentials. 

From Data Access to Business Action 

AI is evolving from answering questions to taking action in business workflows. An answer can inform. An action changes what happens next.  

The latest Autonomous AI Lakehouse capabilities support this evolution by grounding AI agents in business context, enabling multidimensional simulations, and helping govern data consistently across the enterprise data estate. 

These capabilities build on Oracle AI Database’s query optimization, built-in AI, and converged data architecture combining relational data, JSON, vectors, and other data types in a single engine. High concurrency, Exadata performance, and reliability help customers scale analytics and AI while retaining their existing systems, lakehouses, catalogs, and tools. 

Bring AI and the power of Oracle to all your data, wherever it lives. 

Explore Autonomous AI Lakehouse: 

The preceding is intended to outline our general product direction, it is intended for informational purposes only, and may not be incorporated into any contract. It is not a commitment to deliver any material code or functionality, and should not be relied upon in making purchasing decisions. The development, release, timing and pricing of any features or functionality described for Oracle’s products may change and remains at the sole discretion of Oracle Corporation.