Managing the Agent Lifecycle at Scale

This talk was presented at WSO2Con North America 2026. Presenter: Malith Jayasinghe, Vice President, AI, WSO2 Traditional software has a development lifecycle. AI agents need something different — they're nondeterministic, context-driven, and continuously evolving. Managing them at scale requires rethinking the entire lifecycle from the ground up. In this talk, Malith Jayasinghe introduces the Agent Development Lifecycle (ADLC) and walks through what it takes to design, build, deploy, and operate AI agents reliably in production. The session covers the core capabilities required at each stage: Evaluation-driven development as a foundational engineering practice, replacing unit tests as the primary quality gate for nondeterministic systems Deep observability and feedback loops that support continuous improvement after deployment Guardrails and runtime controls to keep agent behavior safe and predictable at scale Platform thinking: treating agent management as a unified control plane across the enterprise, not a collection of point solutions A practical look at what production-grade agent operations actually requires — from the team that built WSO2 Agent Manager. View the slides: https://wso2.com/library/conference/2... #AgentLifecycle #ADLC #EnterpriseAI #WSO2Con #WSO2