Oracle launches agentic AI tools for enterprise data management

Oracle has introduced a series of agentic AI advancements to its AI Database platform, aiming to support organisations in developing and deploying AI applications that access and utilise business data securely across various environments.

The expanded Oracle AI Database enables direct integration of agentic AI with both operational databases and analytic lakehouses, granting AI agents access to real-time enterprise information while integrating large language models (LLMs) trained on public datasets.

This provides customers the flexibility to select their preferred AI models, frameworks, open data formats and deployment platforms.

For users operating on Oracle Exadata infrastructure, the new Exadata Powered AI Search offers accelerated queries for managing substantial, multi-step agentic workloads at scale.

Among the latest features is the Oracle Autonomous AI Vector Database, currently in limited release through both free and developer tiers in the Oracle Cloud.

This service allows developers and data scientists to create vector-driven applications via streamlined APIs and a web interface, combining ease of use with established enterprise security and scalability standards.

The company has also introduced the Oracle AI Database Private Agent Factory, which enables analysts and domain experts to construct and deploy data-driven agents using a no-code builder functioning as a container in cloud or on-premises environments.

This approach maintains data privacy by ensuring that sensitive business information does not leave customer control or enter third-party systems.

Another addition is the Unified Memory Core feature that stores context for multiple types of data, including vector, relational, JSON, graph, text, spatial and columnar, in one platform.

This facilitates low-latency reasoning for AI agents by supporting consistent transactions and security protocols within a converged database engine.

In terms of risk mitigation, Oracle’s updated platform includes new measures such as Deep Data Security.

This function enforces end-user-specific access rules for both humans and agents operating on behalf of users, limiting visibility based on individual roles while centralising security controls at the database level. The system enables dynamic updates to access rules as threats evolve.

For customers requiring heightened privacy assurances, Oracle now offers the Private AI Services Container that permits the operation of private instances of AI models without transmitting data beyond organisational firewalls.

This tool also offloads compute-heavy tasks, such as vector embedding generation from the core database infrastructure.

Oracle Trusted Answer Search was announced as a capability designed to reduce errors from LLMs by matching user questions to pre-existing reports using vector search technology instead of relying solely on probabilistic model outputs.

To prevent vendor lock-in and enable broader interoperability, the company has added support for open standards, including native handling of vector data within Apache Iceberg tables through its Vectors on Ice feature.

Additionally, the Autonomous AI Database MCP Server permits external agents secure access without extensive integration or manual configuration.

With these updates now available, customers can build and operate agentic AI applications connected directly to their business data without requiring additional movement pipelines or specialised skills development.

Oracle Database Technologies executive vice president Juan Loaiza said: “By architecting AI and data together, we help customers quickly build and manage agentic AI applications that can securely query and act on real-enterprise data with stock exchange-level robustness in every leading cloud and on-premises.”