.jpg)
Querying data using natural language is no longer a futuristic promise but an operational requirement within organizations seeking to democratize access to information. Traditionally, extracting knowledge from relational databases required a deep understanding of SQL and the underlying schema, creating a bottleneck between business teams and data. The emergence of large language models (LLMs) has driven NL2SQL (Natural Language to SQL) systems, but their adoption in enterprise environments faces challenges of accuracy, hallucinations, and security. Oracle has taken a step forward by natively integrating this capability into its Autonomous Database through the DBMS_CLOUD_AI package, also known as Select AI. This approach, which injects schema metadata directly into the LLM prompt, drastically reduces the generation of incorrect queries and eliminates the need for external middleware.
From a technical perspective, the DBMS_CLOUD_AI architecture allows defining AI profiles that link external providers —such as Cohere or OpenAI— with specific database objects, thus limiting the exposure surface. Implementation includes configuring secure credentials, network access control lists, and creating the profile using the CREATE_PROcedure. Once operational, the system offers four execution modes: showsql (generates the query without executing it, ideal for auditing), runsql (executes and returns data), narrate (makes a second call to the LLM to summarize results in natural language), and explainsql (explains the logic of the generated query). These capabilities are especially valuable for building executive dashboards or conversational assistants that require immediate and understandable responses.
To orchestrate these flows in distributed enterprise architectures, Oracle Integration Cloud (OIC) acts as an abstraction layer. Through the ATP adapter, DBMS_CLOUD_AI functions are exposed as secure REST APIs, allowing business applications —from a CRM to a corporate web portal— to send natural language queries and receive structured data in JSON without direct database access. This integration reinforces governance and facilitates scalability. However, using external LLM providers implies that metadata and even result data cross the public network, requiring rigorous compliance analysis (GDPR, HIPAA). To mitigate this risk, Oracle offers the OCI Generative AI service, which runs inference within the customer's cloud perimeter, preventing sensitive information leakage.
In this scenario, having a technology partner that understands both cloud infrastructure and data governance is crucial. At Q2BSTUDIO we work on artificial intelligence solutions for companies that integrate these NL2SQL pipelines, combining Oracle capabilities with cloud platforms like AWS and Azure. Our approach includes developing custom applications that incorporate AI agents capable of interpreting natural language questions and executing complex queries on relational databases, all under strict cybersecurity controls. Additionally, we complement these solutions with business intelligence services, such as visualizing results in Power BI, so that management teams can make data-driven decisions without relying on specialized technicians.
The convergence of NL2SQL, AI agents, and cloud orchestration is redefining how companies interact with their information. From report automation to creating virtual assistants that respond in real time, the possibilities are enormous. However, the success of these initiatives depends on careful implementation that respects the principles of least privilege and data protection. At Q2BSTUDIO, we combine deep technical knowledge with a strategic business vision to help organizations unlock the full potential of data through custom software, cloud, and intelligent automation.
