Most AI agent demos stop at autonomous task execution. Real-world systems must balance autonomy with identity, governance, fairness, and human accountability.
In this session, I'll share lessons from building Bethuya, a cloud-native, multi-agent platform used to plan, curate, and operate large-scale technical community events. We'll explore how agents collaborate through MCP, use trusted tools and knowledge sources, and remain governed by explicit human approval workflows rather than opaque decision-making. Inspired by production challenges such as cohort curation, scheduling, and operational planning, the architecture combines agent orchestration, secure identity, observability, and human-in-the-loop controls to create systems that users can trust.
Attendees will leave with practical patterns for building secure, auditable, cloud-native agent systems that go beyond chatbot demos and are ready for real-world adoption.