Can Enterprises Trust AI To Build Apps? Oracle APEX Makes The Case

Oracle recently launched its APEX AI Application Generator on Autonomous AI Database Serverless. The new capabilities address a problem that AI coding tools have created for nearly every enterprise, as users now generate working software from prompts, often without IT involvement.

Note: The author and his firm, NAND Research, have in the past provided advisory and other analyst services to Oracle.

Gartner forecasts that developers outside formal IT departments will account for at least 80% of the user base for low-code tools by 2026, up from 60% in 2021. Agentic AI has accelerated this shift.

The productivity gains are real, but AI-generated applications can leave behind substantial amounts of code that must be secured, tested, patched, and maintained for years.

When the employee who prompted the application changes roles, or the model that wrote it is replaced, the organization can inherit code that few people understand. This can become an audit and security risk that compounds with every new application.

Oracle’s APEX AI Application Generator changes what the AI produces from the start. Its generator outputs structured application definitions that existing review, testing, and deployment processes can govern, and the managed APEX runtime handles execution and patching. That design choice has implications well beyond Oracle’s installed base.

How Oracle Governs What AI Generates

The generator builds on APEX, the low-code platform that has been included with Oracle Database for more than two decades. Its centerpiece is APEXlang, a new declarative language that describes an application, including pages, navigation, forms, reports, dashboards, and business logic, in human-readable text.

APEXlang applies the same division of labor to applications. While the AI expresses intent, the platform handles execution, enforces security, and manages upgrades.

The architecture translates into five business outcomes: