Summary
Enterprise AI agents are no longer confined to chat use cases, they now handle sensitive data, call tools and take actions across business workflows. As their autonomy grows, engineering and product leaders must define non-negotiable and enforceable boundaries to protect customers and business processes. This peer roundtable will explore how guardrails for business processes are designed around domain policies, autonomous agentic actions, compliance enforcement, agent behavioral correction, data access and adversarial inputs.
Discussion Points
1.How do leaders determine the scope of work and responsibilities for the agent?
2.When do you need human approval and when do you allow agents to act automatically?
3.How should the design maintain instruction adherence while preventing a sycophantic agent?
4.How should identity and data access restrictions be enforced?
5.How do teams protect agents from prompt injection and adversarial attacks?
Host
Leon Lobo
Director of Engineering @ Oracle
Leon Lobo is a Software Engineering Director at Oracle with 15+ years of experience building large scale ERP applications and finance systems. He leads the development of next generation agentic workflows for Oracle Fusion ERP Payables with a focus on the engineering and product design challenges of deploying AI in critical business processes. Leon is interested in the unique challenges within Enterprise AI and writes and speaks about the engineering and product design patterns needed to help resolve them.
Sign up to receive news and updates on ELC Annual 2026
