Oracle Database 23ai — released to general availability in 2024 — is the most significant Oracle Database release in a decade. The “ai” in the name is not marketing: it signals a fundamental shift in how Oracle positions the database within the enterprise AI stack. For organisations running Oracle Database today, understanding what 23ai introduces is essential for technology roadmap planning. Here are the five capabilities that will genuinely change enterprise data strategy.
1. AI Vector Search: Making Your Database an AI Engine
The most transformative feature in Oracle 23ai is native vector storage and similarity search — the foundational capability that powers AI applications using large language models. Previously, implementing Retrieval-Augmented Generation (RAG) required a separate vector database (Pinecone, Weaviate, Chroma) alongside your transactional database, creating data synchronisation overhead and architectural complexity.
Oracle 23ai stores vector embeddings directly in the same tables as your structured business data. A customer service application can now run a single SQL query that simultaneously performs semantic search on contract text, filters by customer tier (structured data), and retrieves the three most relevant policy clauses — all in one database round-trip. For Oracle EBS and Fusion customers, this means AI-powered search across your historical transactional data without a separate AI infrastructure layer.
2. JSON Relational Duality: One Table, Two Paradigms
Modern application development often requires both relational integrity (for financial data) and document-style flexibility (for product catalogues, user preferences, event logs). Previously, this meant maintaining separate systems or complex middleware. Oracle 23ai’s JSON Relational Duality Views let developers interact with the same underlying relational tables through either SQL or a JSON document API — automatically. The database maintains ACID consistency and referential integrity regardless of which access path an application uses. For development teams building microservices architectures on Oracle Database, this eliminates an entire category of data model compromise.
3. True Cache: Eliminating Read Replicas
Read replicas are a standard pattern for scaling database read workloads — but they introduce replication lag, synchronisation overhead, and additional licensing cost. Oracle 23ai’s True Cache feature provides a fully consistent, automatically invalidated application-tier cache that requires zero application code changes. Reads served from True Cache are guaranteed to reflect committed database state — no stale reads, no cache invalidation bugs, no application logic to manage cache lifetime. For high-read workloads (reporting dashboards, catalogue lookups, reference data), True Cache can reduce primary database load by 70–80%.
4. Property Graphs with SQL/PGQ
Graph databases have gained adoption for fraud detection, supply chain risk analysis, and recommendation engines — but organisations with existing Oracle investments have been reluctant to add a separate graph database to their stack. Oracle 23ai implements SQL/PGQ (Property Graph Queries), an ISO SQL standard extension that lets you run graph traversal queries on existing relational tables without schema changes or data migration. A fraud detection query that traverses transaction relationships, entity connections, and account histories now runs in standard SQL — on your existing Oracle Database, using your existing DBA skills and security model.
5. Microservices-Native Sagas
Distributed transactions across microservices have historically required complex two-phase commit protocols or manual compensating transaction logic. Oracle 23ai introduces native Saga support — a long-running transaction pattern that coordinates business process consistency across multiple microservices using an orchestration model that Oracle Database manages natively. For Oracle-based microservices architectures, this eliminates a category of distributed systems complexity that has historically required custom middleware.
Upgrade Path from Oracle 19c and 21c
Oracle Database 19c remains on Long Term Support through 2027 (with extended support to 2030). Oracle 21c is an Innovation Release that reaches end of life in 2024 for on-premise deployments. For organisations on 19c, the upgrade path to 23ai is direct. On Oracle Autonomous Database (OCI), 23ai features are available immediately without an upgrade project — another advantage of the managed cloud model.
The AI Vector Search capability alone makes 23ai evaluation urgent for any organisation building AI applications on top of Oracle data. The cost of maintaining a separate vector database, managing synchronisation, and duplicating data will exceed the upgrade investment for most enterprise teams within 18 months.
TechnowayIT’s Oracle DBA practice supports Oracle Database 11g through 23ai environments, including upgrade planning, performance optimisation, and AI Vector Search implementation. Contact info@technowayit.com to discuss your database roadmap.