Skip to Main Content
AI4U: Giving AI Agents Real Memory with Oracle AI Database

About This Workshop

Youtube Video

About This Workshop
Most AI systems can answer questions. AI agents are built to do work.

In this hands-on workshop, you will build AI agents for a financial services company called Seer Equity. Their current systems cannot answer simple questions like loan status or pending applications. Instead of returning results, they explain how to look things up. That breaks trust and frustrates clients.

You will build AI agents that connect directly to a loan database and return real answers. When asked about a loan, the agent queries data and responds with the current status, amount, and owner. You will see exactly what the agent did and why it did it.

During the workshop, you will:

Build AI agents that act on data instead of explaining steps

Store and recall client information across interactions

Use prior decisions to guide new ones

Apply company policies stored in the database

Record every action for audit and compliance

You will use Oracle Database and the Select AI Agent framework to define tools, agents, tasks, and teams using SQL. By the end, you will have a working loan processing system built around AI agents that behave consistently and can be trusted.

No machine learning background required. Just SQL and curiosity.

Workshop Info

3 hours
  • What agents are and why they matter: Understand the difference between chatbots that explain and agents that act on your data
  • How agents plan and execute: Trace the agent loop from understanding requests to coordinating tools and completing work
  • The forgetting problem: Experience why stateless AI fails in production and why memory is essential
  • Building the memory core: Create persistent memory with four types (facts, decisions, context, policies) and semantic vector search
  • Safety and control: Implement role-based agents, automated rules, human oversight, and audit trails for compliance
  • Familiarity with the Oracle Database and SQL is helpful but not required
  • An Oracle Login

Other Workshops you might like