▶ Watch ↗21:39
Conference affiliation: Product Management Architect · Oracle · 2026
Jeff Smith builds and explains tools for people working with Oracle databases. His product-management responsibilities span SQL Developer, Data Modeler, SQLcl, REST Data Services and SQL Developer Web—tools for querying, modeling and exposing database data. Through ThatJeffSmith, Jeff also publishes practical guidance that helps developers use those tools.
Jeff brings decades of database-tool experience to that work. His Oracle author biography describes his product responsibilities, and his own biography places database tools at the center of his career. At AI Engineer NYC 2026, he was credited as Product Management Architect at Oracle. His work connects the tools developers use to query data with a newer question: what access should an AI agent receive when it queries on someone’s behalf?
Jeff favors giving an agent access to predefined reports that a team already understands. A user can request information in ordinary language while the report preserves the business meaning of the question. The agent invokes an established operation instead of reconstructing that meaning from the underlying tables.
He places row-level access decisions in the database, before results enter the agent’s context. The database determines which records the requester may see, and the agent receives the permitted results. Permission enforcement belongs in the system holding the data; it should not depend on an agent withholding records it has already received.
These choices address different parts of securing an agent’s access to data: predefined reports constrain what an operation means, while database permissions constrain which records it returns. Jeff approaches the agent as another database client, carrying established reporting practices and access controls into a conversational interface.
▶ Watch ↗21:39
Loading more talks and highlights…