At Oracle AI World, Oracle announced new capabilities for its Autonomous AI Lakehouse, aimed at giving both people and AI agents the business context behind company data, along with tools to run simulations and govern who can see what. The announcement was dated 26 October 2026.
The headline piece is a Business Semantic Layer. It builds reusable models of dimensions, measures, hierarchies and KPIs from catalog annotations, imported SQL annotations and analytic views suggested by AI, and those models feed natural-language querying, Select AI included. Alongside it comes a Business Simulation Engine that embeds Oracle Essbase for what-if analysis, write-back, allocations and dynamic hierarchy rollups, with simulation reports and Excel-based analysis on top.
“Giving AI access to enterprise data is not enough. To make informed decisions and take trusted action, AI agents need to understand the business context behind that data and operate within an organization’s rules and policies”
Çetin Özbütün, executive vice president, Autonomous AI Database Technologies, Oracle, in Oracle's announcement · oracle.com ↗
Underneath, Iceberg support widens: tables can be created, populated, updated and deleted with SQL, and time-travel data products are supported. The AI Data Catalog can mount external catalogs, Oracle and non-Oracle databases and object-store data without copying them, and query them live, including Databricks Unity Catalog, Postgres and MySQL. A Data Lake Accelerator is claimed to run up to three times faster on 100 TB TPC-DS benchmarks, and Lake Cache moves the most-used Iceberg tables into Exadata flash.
One control plane in the catalog applies role-based, object-level access across Oracle, non-Oracle and Iceberg data, for users and agents alike. The platform runs on OCI, AWS, Azure, Google Cloud, Exadata Cloud@Customer and OCI Dedicated Region, and a Data Studio workspace needs no database provisioning. Liberty Energy, a North American oil and gas services firm, uses it to unify financial, operational, supply chain and workforce data.
Stock illustration used with InfoWorld's report on Oracle's lakehouse additions
Shutterstock · Image: InfoWorld
Nipun Agarwal told theCUBE that the lakehouse has become the "decision backbone" for AI, pitched as an open layer across existing data rather than a replacement. He said Oracle does not expect to be the only lakehouse enterprises use, and argued that real-time access matters more when the consumer is not necessarily human, since scheduled copies fall short.
About the Oracle Autonomous AI Lakehouse
Oracle Autonomous AI Lakehouse is Oracle's open data platform, combining Oracle Autonomous AI Database with the Apache Iceberg standard. Oracle introduced it at Oracle AI World on 14 October 2025, and it runs on OCI, AWS, Azure, Google Cloud and Exadata Cloud@Customer.
This week's update builds on that launch. At the start it added native Iceberg access through catalogs including Databricks Unity, AWS Glue and Snowflake Polaris, alongside an Autonomous AI Database Catalog described as a "catalog of catalogs". Oracle also unveiled Data Lake Accelerator then, which scales network and compute for large Iceberg queries and bills only for the resources used.
Analysts quoted by InfoWorld expect business metrics to make agent output more reliable and planning faster. They also warn that companies must define and maintain their metrics, relationships and rules, that incomplete models can still mislead, and that catalogs and semantic layers risk lock-in. In planning software Oracle faces Anaplan, OneStream and SAP Analytics Cloud. The semantic layer and catalog capabilities are due within twelve months; Oracle gave no date for the simulation engine or the expanded Iceberg support.
Sources · 3