Oracle makes a key database for agent AI development

Enterprise AI deployments are stalling not because agents are difficult to build, but because organizations lack the data infrastructure to run them reliably at scale. The shift from chatbots to autonomous, multi-step agents has exposed a structural gap in agent AI development.

Oracle Corp. positions the database as the main center of gravity for business agent AI, arguing that the future of intelligent applications will not be determined by model performance alone, but by how deeply AI is integrated into the underlying data layer. That conviction is now being translated into concrete design, according to Tirthankar Lahiri (pictured), senior vice president of critical information and AI engines at Oracle.

“Agency systems will be the future of application development. They are the present and the future,” Lahiri told CUBE. “Many organizations still struggle to get the most out of agents, because ultimately, agents are only as good as their data.”

Lahiri spoke with CUBE’s Dave Vellante at the Oracle Data Deep Dive NYC event, during an exclusive broadcast on CUBE, SiliconANGLE Media’s live streaming studio. They discussed Oracle’s strategy for embedding agent AI development and the company’s approach to AI data security and open standards. (* Disclosure below.)

Agent AI development focused on integrated memory and data layer

Oracle’s approach challenges the prevailing assumption that agent AI is primarily an orchestration problem. Rather than represent a separate agent layer that sits on top of separate data stores, Oracle collapses the stack – running the agent logic as close to the data as possible. The company’s AI Database Private Agent Factory and Autonomous AI Vector Database illustrate that thesis, giving developers and business users alike a seamless way to build and deploy agents against live business data without moving it between systems, Lahiri explained.

“There are two types of agents. There are logic-oriented agents and data-oriented agents,” he said. “Data-centered agents work best when integrated with data. We want to eliminate the need for multiple round-trips – accessing multiple databases. Designing agent processing and data access avoids fragmented or broken AI. You get AI that works on clean, real-time, current data without the need to separate your data into multiple repositories.”

Key to Oracle’s design is what the company calls the Unified Memory Core – an ability that gets the full type of agent’s memory, from short-term context to long-term true organizations, from a single unified data store, Lahiri explained. Rather than routing agents to separate a graph, document or vector database for different processing tasks, Oracle allows a single underlying data layer to handle all of those needs simultaneously. This development of agent AI eliminates the overhead synchronization and consistency risks that come with managing multiple specialized systems.

“Sometimes you want associations and you want a graph of information. Sometimes you just want a true representation of the event that happened,” he said. “That output, when done locally with real data, is current, consistent and fully secure. We call that Unified Memory Core for that reason, which is more efficient than using multiple storage systems to represent different types of memory.”

The same approach to data extends to Oracle’s approach to AI data security. As agents move from answering questions to taking action – performing tasks, accessing sensitive records, running business processes – security enforced at the application layer becomes inadequate. Oracle’s answer is what it calls serious data security: the implementation of a policy embedded directly in the database, ensuring that even a dynamically generated or counter-injected query cannot return data that an authorized user is not authorized to see.

“The problem we have today is in many programs, security is built in the application phase,” said Lahiri. “The only way to solve this problem is to get the data from the source. Even if the query is not well-formed, you can’t get back data that you shouldn’t show. That’s really what deep data security gives you – and I think in this world of AI, that’s the only way to protect data.”

Stay tuned for the full video interview, a SiliconANGLE and CUBE segment covering the Oracle Data Deep Dive NYC event.

(* Disclosure: TheCUBE is a paid media partner of the Oracle Data Deep Dive NYC event. Neither Oracle, the CUBE’s event sponsor, nor other sponsors have editorial control over the content on The CUBE or SiliconANGLE.)

Photo: SiliconANGLE