Select AI gives developers, analysts, and application teams a natural-language layer over enterprise data, so they can generate SQL, chat with data, ground responses with trusted content, create synthetic data, and now build agentic workflows directly in the database. Oracle’s current Select AI capability matrix shows that the feature set is available across both Oracle Autonomous AI Database and Oracle AI Database, with some advanced capabilities concentrated in the newer 26ai releases.

Just as importantly, Select AI is included with Oracle AI Database and Oracle Autonomous AI Database. You bring the AI models you want to use, choosing from the wide range of supported providers and models that fit your requirements. Oracle documentation and product collateral emphasize this flexibility, noting support for multiple providers, including OCI Generative AI, OpenAI, Azure OpenAI, Cohere, Google, Anthropic, Hugging Face, Amazon, OpenAI-compatible providers, and private LLM deployments, depending on release.

The platform view at a glance

Across both Oracle Autonomous AI Database and Oracle AI Database, Select AI provides a consistent core experience in both 26ai and 19c, including chat, natural language to SQL (NL2SQL) generation, synthetic data generation, AI agents, summarization, and translation.

The 26ai releases build on that common foundation with added capabilities for NL2SQL feedback and auto object selection, and retrieval-augmented generation (RAG). These features rely on Oracle AI Vector Search, which is not available in 19c.

For Oracle AI Database, the version number 23.26.1 corresponds to the 26ai release.

That means organizations standardizing on 26ai get the expanded Select AI experience, while 19c still delivers a strong baseline that includes chat, NL2SQL, synthetic data generation, AI agents, summarization, and translation.

Key Select AI capabilities

NL2SQL

Natural language to SQL is a core capability of Select AI, which enables users to describe the data result they want in natural language and then generate, run, narrate, or explain the resulting SQL. Select AI uses schema metadata to produce an augmented prompt that is use by the LLM specified in your AI profile. For developers, NL2SQL generation accelerates query authoring, and for business users, it opens access to enterprise data without requiring deep SQL expertise.

Feedback

In 26ai, Select AI adds Feedback for NL2SQL, allowing users to provide feedback on generated results from the SQL command line or via PL/SQL procedures. The goal is straightforward: improve query generation accuracy over time by learning from user corrections and preferences. That makes Select AI more adaptive in real-world enterprise usage, where schema complexity and business language often evolve together.

Auto Object Selection

Also in 26ai, Select AI for NL2SQL can automatically select the relevant table metadata for a prompt and send only the needed metadata to the LLM. This uses a vector index automatically created over the objects specified in your AI profile. Based on the user prompt, semantic similarity search help identify relevant metadata for the query, which helps simplify prompt construction and improve accuracy by narrowing context to the most relevant objects.

RAG

Select AI retrieval-augmented generation combines LLM reasoning with trusted enterprise content. Oracle automates the RAG flow from generating embeddings to retrieving semantically similar content from vector stores and augmenting prompts with that data. The result is more relevant, grounded answers with lower hallucination risk, especially for private, domain-specific information.

SDG

Synthetic data generation helps teams create schema-conformant data for development, testing, UX validation, and machine learning projects without exposing sensitive production data. Use cases include populating metadata clones, jump-starting new projects when real data is unavailable, and supporting AI/ML use cases where data access is restricted.

AI agents

The Select AI Agent framework extends the platform from prompt response to action-oriented workflows. This is an autonomous agent framework for building and managing agents inside the database, where agents can reason about requests, call built-in or custom tools, reflect on results, and maintain short- and long-term memory across conversations. In practice, that means developers can create agentic workflows using SQL and PL/SQL while keeping logic, governance, and data close together.

Summarization

Summarization is one of the most immediately useful everyday capabilities. Using your specified LLM, Select AI can return text responses, summarize query results in prose, and enable conversational follow-up, which makes it easier to turn database output into readable business language for reports, dashboards, and embedded applications.

Translation

Translation extends the value of natural-language interaction by helping teams generate multilingual responses and adapt content for global users. Users can use translations services from OCI, GCP, AWS, and Azure.

Chat

Chat provides the basic interaction model for Select AI, allowing users to submit prompts ranging from simple questions to complex requests. Those prompts are passed directly to the LLM, unaltered, so the user’s original intent is preserved.

Conversations, which applies to all Select AI capabilities, build on that foundation by supporting multi-turn exchanges between you and your LLM, maintaining context across prompts and enabling more natural follow-up interactions.

Why 26ai matters

The jump to 26ai is significant because it turns Select AI from a strong natural-language interface into a more complete AI application layer.

  • Feedback improves NL2SQL accuracy over time
  • Auto object selection reduces manual schema targeting
  • RAG grounds responses in trusted enterprise content

Combined with AI agents, those additions make 26ai the release where Select AI becomes especially compelling for enterprise-grade AI applications and workflows.

Closing thought

The message is simple: Select AI is available across both Oracle Autonomous AI Database and Oracle AI Database, with 26ai delivering the broadest feature set. And because the AI models are supplied separately, you can pair Oracle’s in-database AI framework with the provider and model strategy that best fits your performance, governance, and cost requirements.

Resources

For more information…