Oracle Targets the Business Definitions Enterprise AI Gets Wrong

Key Highlights

• Oracle is addressing a practical enterprise AI problem: accessing the right data does not guarantee interpreting it correctly.
• A shared semantic layer could make approved business calculations reusable across AI applications.
• Expanded Iceberg support gives Oracle a broader role in existing lakehouses without requiring wholesale data migration.
• Integrating simulation could help teams evaluate proposed changes before approving business actions.

The News

At Oracle AI World on October 26, Oracle announced updates to Autonomous AI Lakehouse spanning business semantics, Apache Iceberg, governance, and simulation. The company plans to deliver the Business Semantic Layer and agentic AI Data Catalog capabilities within the next 12 months. The release also describes expanded SQL operations on Iceberg tables and an Essbase-based Business Simulation Engine.

Analyst Take

An AI agent can query the correct table and still produce the wrong business answer. “Revenue” might mean booked revenue, recognized revenue, or cash collected. A technically valid query can use the wrong definition, apply the wrong reporting period, and return a number that looks entirely reasonable. Giving an agent access to more data does little to resolve those differences. Oracle’s proposed semantic layer addresses a more useful question: which business definitions should the application use?

This is a useful place for Oracle to invest. Business definitions have a direct effect on whether an AI application produces an answer an enterprise can use. Making approved calculations reusable could save developers from rebuilding the same logic across applications and give business teams a clearer way to review it. When a calculation changes, teams could update the shared definition rather than track down separate implementations. That is a concrete improvement to the AI stack, even if it is less attention-grabbing than a new model.

Enterprises will still need to test whether an application chooses the right definition. A company may have several legitimate revenue measures. The application needs to recognize when the question is ambiguous and ask for clarification rather than quietly selecting one. Someone also has to own those definitions and resolve disagreements between departments. AI-generated descriptions can help document the data, but business owners need to approve what those descriptions mean.

Oracle is entering an established competitive area. Microsoft, Snowflake, and Databricks already give enterprises ways to define and reuse business metrics across their analytics environments. Oracle’s strongest case is likely to be with customers that want to extend existing Oracle database and analytical investments across a broader data estate. Enterprises with established semantic models elsewhere will need a clearer reason to add another implementation, particularly if their employees already rely on those definitions through existing reporting and AI tools.

Interoperability will help determine that value. Oracle describes importing compatible semantic models through Apache Ossie, the initiative formerly known as Open Semantic Interchange. That could reduce the need to recreate business definitions, although connecting to an external catalog is not the same as preserving all its calculations and policies. Customers should test what survives an import, whether results remain equivalent, and how changes are maintained. Oracle’s opportunity is to make existing business definitions useful across more workloads without creating another set that teams must reconcile.

Bringing simulation into the lakehouse could give AI applications a useful intermediate step between answering a question and recommending an action. A proposed purchasing change, for example, could be evaluated against inventory and cost assumptions before someone approves it. That is a practical use of established planning technology alongside AI. The opportunity is to help employees examine a recommendation’s consequences while keeping approval tied to the business process.

What Was Announced

Oracle’s proposed Business Semantic Layer uses annotations and reusable analytic views to represent metrics, hierarchies, and calculations. The agentic AI Data Catalog is intended to enrich metadata and help identify relevant data across connected sources. Both carry a delivery window of the next 12 months.

Oracle’s technical overview describes connections to Databricks Unity Catalog and Snowflake Horizon Catalog, alongside support for importing compatible semantic models. These connections could help customers preserve existing work rather than recreate every definition within Oracle.

Expanded Iceberg functionality covers creating, populating, updating, and deleting tables through SQL, with time travel supporting investigation and recovery. Data Studio provides a workspace for preparing and analyzing data. The Essbase-based simulation engine adds multidimensional planning and what-if analysis, including write-back and Excel-based analysis.

Oracle also describes object-level access policies and temporary, table-scoped credentials for supported Iceberg access. Its technical overview states that catalog policies operate alongside underlying source controls. These are controls over data access; applications still need to establish which business actions an agent is authorized to perform.

The announcement therefore spans several kinds of work: finding relevant data, interpreting it, preparing it for analysis, and modeling potential decisions. Customers should evaluate those capabilities against a specific workflow and confirm the availability of the components it requires.

Looking Ahead

We believe that the next breakthrough in enterprise AI isn’t simply about accessing more data, it’s about understanding what that data means and putting it to work. With Autonomous AI Lakehouse, Oracle is attacking both challenges head-on. Its Business Semantic Layer gives AI agents a shared understanding of business metrics, relationships, and KPIs, while expanded Apache Iceberg support brings the power of Oracle’s analytics and AI directly to open enterprise data. This is a powerful combination that bridges the gap between raw data and business intelligence, positioning Oracle to turn the open lakehouse into a foundation for business-aware agentic AI. The real game changer is making open data not just AI-accessible, but AI-intelligent.

The 12-month delivery window matters because the semantic layer is central to this announcement’s AI value. As it becomes available, customer evidence should show whether teams can reuse existing definitions, maintain them across connected sources, and improve answers to actual business questions. The useful measures will include how much application logic developers can reuse and how often employees still need to reconcile conflicting results.

Integration with existing catalogs will deserve close attention. Importing a definition provides a starting point. Customers also need to know who owns it, how updates propagate, and which applications are affected when it changes. Those maintenance details will help determine whether a shared semantic layer reduces work or creates another set of definitions to manage.

We would look for applications that recognize ambiguity as well as calculate correctly. If an employee asks for “revenue” without specifying which measure, asking a follow-up question may be the most accurate response. Oracle’s opportunity is to make that business discipline easier for developers to build into their applications.