In just a few years, AI has gone from novelty to ubiquity, shaping everything from how we find information to how we work. For consumers, AI is very accessible and effortless: ask a question, get an answer; draft a document; generate a travel plan; create images or videos, and more.

Many organizations, however, are still stuck in pilot mode while others are seeing AI projects fail in production. The challenge is the data foundation. While LLMs are exceptionally good at using the world’s public information, they need trusted access to private business data to generate business value. This data is usually scattered in silos across formats, systems, and clouds. As a result, developers and IT teams must build patchwork solutions, often leading to challenges including performance bottlenecks, brittle integrations, complex pipelines, and unreliable answers. At the same time, AI is evolving from chatbots to AI agents that can autonomously plan, make decisions, and execute multi-step actions. But for Agentic AI to be effective, it requires fast and secure access to real-time operational data, which makes the data foundation even more critical.

With 97% of Fortune Global 100 customers trusting Oracle for their business data, we understand the importance of activating this data for AI innovation without compromising security.  This is why we are architecting AI directly into Oracle AI Database, so it’s simple to use, fast, and secure. In fact, customers such as Munich Re HealthTech, Rappi, Retraced, and Uniti are already benefiting from the AI capabilities in Oracle AI Database.  

Today, we’re excited to introduce agentic AI innovations that transform how organizations like yours build, deploy, and scale AI safely and securely. Now you can easily use your business data with LLMs trained on public data to deliver impactful results, taking advantage of capabilities that help you innovate faster with AI designed for data, minimize AI data risk, and end AI data lock-in with open standards and frameworks.

Innovate Faster with AI Designed for Data

Oracle AI Database enables you to build trustworthy agentic AI applications designed to maximize the value of your business data with minimal effort. With agentic AI capabilities integrated directly into a secure, scalable database, there is no need for you to develop complex data-movement pipelines that delay outcomes.

Oracle AI Database provides a unified memory core for AI agents to store context and session information. AI agents must often process vector, JSON, graph, columnar, spatial, text, and relational data in the same context, but only Oracle AI Database processes these diverse data types in the same engine, with the same mission-critical transactional performance, high availability, and robust security. With a database engine designed to concurrently support transactional and analytical processing, agents can reason over all your data, including live business data, with minimal latency. At the same time, this eliminates the complexity and staleness of syncing external data stores.

To help developers quickly build apps that use AI search, we introduced Autonomous AI Vector Database, a fully-managed vector database service purpose-built to simplify and accelerate AI adoption. It enables developers to quickly build vector-powered applications using intuitive APIs and an easy-to-use web interface. Built on Oracle Autonomous AI Database’s trusted data management and advanced capabilities, it combines a streamlined developer experience with enterprise-grade security, reliability, and operational control. Currently in Limited Availability, Autonomous AI Vector Database will be accessible through the Oracle Cloud Free Tier. In addition, a Developer Tier with low-cost pricing will be available. Both tiers will have a one-click upgrade to Autonomous AI Database to accommodate growth in processing requirements. Read the technical blog to learn more.

AI Database Private Agent Factory is another new innovation: a no-code platform that enables business analysts and domain experts to quickly build, scale, and safely deploy agents and workflows. The Private Agent Factory framework runs as a container in public clouds or on-premises, maintaining data security by enabling you to develop and orchestrate AI agents without having to share data with any third-party. An AI agent builder with a visual interface and a template library makes it simple to take advantage of the full power of Oracle AI Database in creating and managing intelligent data-centric agents. Private Agent Factory includes several pre-built data-centric AI agents such as a Database Knowledge Agent, a Structured Data Analysis Agent, and a Deep Research Agent. Developers can also use the Select AI agent framework to build, deploy, manage AI agents, and customize pre-built agents within Autonomous AI Database, using familiar PL/SQL or Python.

Watch this Private Agent Factory demo to see it in action and read the technical blog to learn more.

Minimize AI Data Risk

New AI applications raise the stakes for data security and trust. The risks range from inaccurate outputs to the potential exposure of private and sensitive data. In many applications today, fine-grained, end-user level access controls are enforced in application code rather than in the data layer. As a result, apps commonly connect to the database using privileged accounts rather than end-user credentials.  This approach is difficult to verify or enforce, as controls are often bypassed by different tools, and are difficult to keep consistent at scale because they must be enforced in every application that accesses the data. AI amplifies the risk of sensitive data exposure as agents can generate a wide range of SQL queries. AI also introduces new threats such as prompt injection, where users may try to manipulate the model to bypass guardrails and gain access they should not have in the first place.

Oracle AI Database provides industry-leading data security and role-based access controls at the data layer, preventing unauthorized access to sensitive business data by LLMs and users alike. To extend the robust security capabilities of Oracle AI Database, we are introducing Oracle Deep Data Security to protect against new AI-era threats, such as prompt injection. Deep Data Security uses declarative, database-native controls that enforce end-user access privileges at the row and column level. By centralizing and decoupling end-user security from application code, it enables organizations to continuously update defenses as new threats emerge. Read the technical blog to learn more.

Since probabilistic LLMs can occasionally hallucinate, enterprises can instead rely on Trusted Answer Search when answers must be deterministic.  It uses AI Vector Search rather than an LLM to find the best matching resource in response to natural language questions. APEX AI Interactive Reports provides trusted answers restricted to data in a pre-defined data view. An LLM converts natural language questions into filters and aggregations that can be easily understood and edited by the end-user. Millions of existing APEX Interactive Reports will be AI-enabled by simply upgrading to the upcoming version of APEX.

Another risk to consider is the effect of AI agent workloads on back-end databases’ performance, scalability, and availability. As your organization is set to give tens of thousands of autonomous AI agents access to enterprise data on top of the existing thousands of users, back-end databases will need to process millions of additional transactions. AI agents will place unprecedented demands on data infrastructure, requiring agent parallelism for high-volume interactions, AI vector search across massive data sets, and transactional capabilities on back-end systems.

Oracle Globally Distributed AI Database, powered by Exadata, is highly performant and scalable on demand to handle peak workloads—with up to 5 exabytes of storage capacity and over 100,000 CPU cores. In addition, agentic workloads need always-on unified access to distributed data. Oracle AI Database, specifically Globally Distributed Exadata Database on Exascale Infrastructure, is designed to address the stringent availability and data residency requirements of agentic AI workloads.

End AI Data Lock-in with Open Standards and Frameworks

We aim to provide flexibility and choice whenever possible and help you avoid lock-in. Over the last several years, we have made Oracle AI Database available in every leading cloud and on-premises so you can choose where your data and AI workloads run. But flexibility goes beyond deployment locations. To help end data lock-in, Oracle AI Database gives you choice across the full stack: your preferred cloud or platform, your preferred AI model, and your preferred application-tier agentic framework. You can build and run agentic AI apps using open standards, languages, APIs, and data formats.

One increasingly important open data format is Apache Iceberg. Oracle Vectors on Ice now provides native support for vector data in Apache Iceberg tables and object storage. It can read vector data directly from Iceberg tables, create vector indexes in Oracle Autonomous AI Vector Database, and automatically refresh those indexes as the underlying vector data changes in the Iceberg tables. Oracle AI Database also enables unified hybrid vector search on Iceberg tables. Vector data stored in file formats such as CSV, text, JSON, ORC, AVRO, Parquet, and others can also be read into Oracle AI Database for indexing and search. As a result, workloads with both unstructured data and vectors in object storage can benefit from the performance, scalability, and security of Oracle AI Database. Read the technical blog to learn more.

That same commitment to openness extends to the agent standards and frameworks that you rely on. Oracle also introduced Open Agent Specification, an open, framework-agnostic standard for defining AI agents and workflows in a unified declarative format. It enables portability, reusability, and interoperability across platforms, including easy import and export of agents built with AI Database Private Agent Factory or Select AI Agent. Oracle AI Database is also integrated with popular application-tier agentic frameworks, including OCI Generative AI, Google Vertex AI, Amazon Bedrock, LangChain, and CrewAI, so you can build with the tools and frameworks you already use.

Additionally, Oracle continues to embrace MCP, which is a rapidly emerging open standard for connecting AI models with enterprise tools and data. Oracle SQLcl MCP Server for Oracle AI Database, available via the Oracle SQL Developer VS Code extension for Oracle Database 19c and later releases, provides a secure, easy way to connect AI to Oracle AI Database. Autonomous AI Database MCP Server is a fully-managed feature that lets external AI agents and MCP clients securely access Autonomous AI Database and its capabilities without custom integration code or manual security administration.

Together, these open standards, integrations, and deployment options give you the flexibility to build, deploy, and run AI agents across OCI, AWS, Microsoft Azure, Google Cloud, as well as in hybrid and on-premises environments.

Build What’s Next with Agentic AI and Your Business Data
The next wave of enterprise AI will be defined by what your organization can do securely with your business data in production. With Oracle AI Database, you can build agentic AI applications faster, scale with confidence, and deliver real business value with the performance, security, and flexible, open approach that businesses require in real-world production environments.

While other approaches rely on external agent orchestration or must make calls to different types of databases, Oracle has simplified agentic AI for business users by architecting it directly into Oracle AI Database, providing consistency and simplicity with the same security, resiliency and scalability for every agentic workload.

Many of the Agentic AI capabilities built into Oracle AI Database are immediately available to organizations and developers worldwide, enabling you to start developing and deploying game-changing agentic AI applications without moving data, learning new skills, or struggling with database scalability. With these new innovations, your organization can move from experimentation to production faster, and build what’s next using agentic AI with your business data.

More Information

In addition to the innovations discussed in this blog, you can explore a few others: