What should change and what should stay as data architecture evolves for analytics and AI

At a glance: With Oracle Autonomous AI Lakehouse, organizations can run analytics and AI across Oracle and non-Oracle data while continuing to use the systems their businesses rely on. It supports three common starting points: modernizing an on-premises Oracle data warehouse, extending Oracle Autonomous Database, or connecting Oracle data with data stored elsewhere.

Enterprises rarely redesign their data architecture from a blank slate. Existing databases, applications, etc. and other systems , etc. already support critical operations. New requirements for analytics, AI, and business applications increase the demand for timely, relevant information from systems that often were not designed to work together.

Yet modernization discussions often begin with a new destination: move or duplicate data, rebuild pipelines, and recreate applications and controls on another platform. That may be appropriate for some workloads, but across an entire data estate, it can mean rebuilding business logic, reporting, security policies, and processes that already work. This can add cost, complexity,  and implementation time. Extending what already works allows organizations to focus investment on the capabilities they need next while continuing to use established platform capabilities, applications, and processes.

AI makes these tradeoffs harder to ignore. An AI application working with incomplete, stale, or poorly understood data can produce inaccurate answers. Gartner found that 63% of organizations either lack or are unsure whether they have the right data-management practices for AI.¹ But modernization should begin with the business outcome the organization needs, whether that is supporting a new application, improving analytics and warehouse performance, or preparing data for AI.

Build on the data estate you already trust

Existing Oracle environments contain years of business logic, security policies, and operational knowledge. Modernization should build upon those assets while allowing the architecture to evolve according to the requirements of each workload.

Oracle Autonomous AI Lakehouse (AI Lakehouse) is the next generation of Autonomous Data Warehouse. It retains existing data warehouse functionality while adding new capabilities for working with distributed data. Native support for Apache Iceberg™, an open table format, allows teams to query tables in object storage, including tables managed through supported external catalogs. AI Lakehouse also includes Data Studio for defining shared business semantics through common semantic models, Select AI for asking questions in natural language, and AI Vector Search for finding relevant information based on meaning. Because it builds on the Oracle Database foundation customers already use, organizations can continue using the applications, SQL, data models, security controls, and business processes that depend on it.

Omdia Senior Analyst Stephen Catanzano described Oracle’s approach as bringing AI to where the data already lives and avoiding the “data gravity tax” associated with lift-and-shift strategies.²

How does AI Lakehouse support different environments?

The right first step depends on the systems your organization already has in place and the business requirement being addressed.

How can organizations find and query data across systems?

Extending existing systems requires interoperability to create a common view of enterprise data while allowing the catalogs that serve individual platforms to remain in place.

Oracle AI Data Catalog uses a catalog-of-catalogs architecture to connect metadata from Oracle sources and supported external catalogs, including AWS Glue, Databricks Unity Catalog, and Snowflake Horizon Catalog. Those catalogs can continue serving their existing platforms while Oracle AI Data Catalog gives your teams a shared view of data across the organization, including its location, lineage, technical characteristics, and business meaning.³

AI Lakehouse supports multiple ways to query distributed data. These include catalog-based access to data exposed through connected catalogs, federated queries against supported databases, and native Apache Iceberg support for tables in object storage.⁴

Query patterns also differ. Lake Cache stores frequently accessed external data locally within AI Lakehouse, which can improve performance for repeated queries against external tables and catalog-backed data. Data Lake Accelerator temporarily adds compute capacity for large scans of external data in object storage and releases those resources when the query finishes. These capabilities can help improve query performance without requiring your teams to maintain another permanent copy solely for analytics.

During early testing, SKY Brazil reported improved query speeds on external data in object storage with Data Lake Accelerator. On-demand scaling allowed the team to handle complex queries as needed while keeping costs under control.⁵

Reducing complexity across applications and workflows

Modern analytics and AI applications often need to combine transactional data with documents, vectors, spatial information, and relationships among entities. When each data type requires a separate engine, developers must integrate multiple platforms, security models, and query results before an application can use them together.

AI Lakehouse uses Oracle’s converged database engine to support relational, JSON, graph, spatial, vector, and XML data on a common foundation. These workloads share the same transaction manager, optimizer, and data access controls, and developers can work across them using SQL. Reducing the need to integrate separate engines can help simplify implementation and reduce the number of separate solutions that your organization needs to integrate and manage.

AI Lakehouse also can reduce the number of separate tools needed for data preparation, analysis, and AI development. Data Studio supports data loading, transformation, analysis, and sharing. Its Business Models tool creates shared business semantics by allowing teams to define hierarchies, dimensions, measures, and calculations once and reuse them across analysis applications. Select AI translates natural-language questions into SQL, while AI Vector Search finds relevant information based on meaning rather than exact keyword matches.⁶

Protecting data used by applications and AI agents

Applications, analytics tools, and AI agents can all access the same enterprise data. When each access path implements its own authorization rules, inconsistencies expose information that a user or agent should not have access to.

AI Lakehouse applies access policies at the data layer. Oracle Deep Data Security enforces those policies directly in the database, helping apply defined access policies to uses and AI agents, including when an agent acts on a user’s behalf. Because enforcement remains with the data, the same rules can apply across applications, analytics tools, and AI agents.⁷

Autonomous operations handle routine provisioning, tuning, patching, backup, and scaling. Autonomous Data Guard provides automatic failover, and eligible AI Lakehouse deployments with it enabled receive a 99.995% availability SLA.⁸

What does AI Lakehouse look like in practice?

Begin with the environment and business priority your organization has today. These examples illustrate the three starting points.

• Modernize an on-premises data environment. Liberty Energy, a North American energy services company, uses Oracle Autonomous AI Lakehouse to bring together financial and operational data from cloud and on-premises systems. Built with Oracle AI Data Catalog, OCI Object Storage, and Apache Iceberg, the platform integrates more than 50 TB of data from enterprise applications, object storage, and databases across AWS and Azure. The company reduced manual reporting and reconciliation effort by 30% while supporting more than 700% growth.
• Extend an existing Autonomous Database deployment. Bitron, a global electronics manufacturer, built on its existing Autonomous Data Warehouse deployment with AI Vector Search. It combined fault tickets, technical manuals, and charging-station telemetry in a conversational assistant that now answers 85% of operator questions and helps diagnose issues five times faster, helping the company reduce downtime and operating costs.
• Connect Oracle data with other enterprise systems. Mars Veterinary Health connected Oracle data with approximately a dozen non-Oracle finance, procurement, and supplier systems. Data that previously refreshed every four to six hours now refreshes every 30 minutes, helping the organization close fiscal reports within two days and giving finance teams more time for analysis.

Turn your existing data estate into a foundation for AI

Start with the business outcome that matters most, then determine which data and workloads need to change to deliver it. Oracle Autonomous AI Lakehouse lets you take that next step while preserving the data, logic, and controls your business already depends on.

• Explore Oracle Autonomous AI Lakehouse and find the starting point that fits your organization.
• See Oracle Autonomous AI Lakehouse in action

1. Gartner Press Release, Lack of AI-Ready Data Puts AI Projects at Risk, February 2025. GARTNER is a trademark of Gartner, Inc. and/or its affiliates. ↩︎
2. Oracle, “Leading Analysts Highlight the Advantages of Oracle Autonomous AI Lakehouse” ↩︎
3. Oracle Autonomous AI Lakehouse ↩︎
4. Oracle documentation, “Use Lakehouse with Autonomous AI Database” ↩︎
5. Oracle, “Autonomous AI Lakehouse Enables Open, Interoperable Data Access Across Multi-Platform, Multicloud Environments” ↩︎
6. Oracle Autonomous AI Database Data Studio features ↩︎
7. Oracle, “What Is Oracle Deep Data Security” ↩︎
8. Oracle documentation, “Availability Service Level Agreements for Autonomous AI Database” ↩︎