SiliconANGLE theCUBE · Nov 5, 2026 · 19m · 11m saved
With Hasan Rizvi, Executive Vice President of Oracle Database Engineering
A generic AI assistant asked why an order is late answers with a tracking link. Hasan Rizvi, executive vice president of database engineering at Oracle, told Dave Vellante on SiliconANGLE theCUBE that an assistant connected to the company's own data can say where the stock is and how to speed up delivery.
Most of the AI debate is about which model is smartest. Hasan's case is that the model matters less than the data it can reach.
"If you ask a generic AI assistant, you get a generic answer, check the tracking link. If you ask an AI assistant that knows your business, you get a much better answer." — Hasan Rizvi
Hasan leads engineering for a database engine Oracle has worked on for 40 years, and he was presenting Oracle's AI plans for it at Oracle AI World.
The full interview is covered here so you can skip it. 19 minutes of audio, 8 minutes of reading.
Here are the 9 takeaways that matter.
Key Takeaways
Data, not the model, is what makes AI useful to a business, Hasan says
An AI model that can write SQL still does not understand the business, so Oracle defines business terms in the database
An AI agent needs data, memory, tools, policies and permissions, not only a model
Data chaos is built one database at a time, and AI will make it everyone's problem
Oracle keeps customers' choice of where to run, how to manage, which AI and which data format
AI has made it easy for attackers to reverse-engineer security fixes
1. Data Makes AI Useful
Hasan built his Oracle AI World session around the late-order example. His argument was that the data a company holds is what turns a general-purpose model into something that can run a business.
Hasan Rizvi says data is what makes general AI useful to a company
"Data is what makes general intelligence into a intelligent business." — Hasan Rizvi
Oracle Database, he said, is built to sit where AI meets a company's most important data and transactions. He set out four goals for it: AI designed for data, an end to data chaos, an end to data lock-in and an end to data risk.
