Oracle performance optimization has never been a discipline for "button pushers". If you want to understand database instances, system design, or stubborn SQL statements, you need experience, measurements, and a sense of which lever matches the problem. AI does not change that foundation, but it can make the work much faster. If you use it right!

This session shows how experienced DBAs and database developers can use AI as a tool for better Oracle performance: reviewing large data sets, evaluating AWR reports, and working through execution plans. As a human, I want to prepare hypotheses and back them up with evidence; the machine helps me to recognize patterns.

The central point: a human using AI works more efficiently than a human without it, and far better than a machine being let loose on a "real" performance problem on its own. Judgment remains craftsmanship: understanding context, checking measurements, identifying side effects, and assessing risks. I would also rather talk to the user myself...

We will also discuss limits: data protection, hallucinations, context windows, poorly connected sources, and unwanted change through MCP. Magic prompts will not solve that. Old DBA virtues will: validation, guardrails, least privilege, test cases, observability, responsibility, reproducibility.

For experienced DBAs and database developers. Focus: SQL query optimization, complemented by know-how about instance and system.