Managing the Agent Lifecycle at Scale
As enterprises move from AI experimentation to production, building and operating AI agents reliably at scale introduces new challenges. Unlike traditional software, agents exhibit non-deterministic, context-dependent behavior. They can decide how to approach a task, which tools to use, and what actions to take. This requires rethinking the traditional software development lifecycle and extending it into an Agent Development Life Cycle (ADLC).
In this talk, we'll explore the key aspects of the ADLC, from defining and designing agents to evaluating their behavior, observing their execution, enforcing runtime boundaries, and governing their interactions with LLMs. We'll then examine how these challenges evolve as agent adoption grows across teams, frameworks, models, and environments. Finally, we'll make the case for an Agent Control Plane as the foundation for managing the agent lifecycle consistently at enterprise scale.
Speakers
Malith Jayasinghe is VP of AI at WSO2, where he leads initiatives to build scalable, secure, and production-ready AI systems for the enterprise. With over 15 years of experience in building, scaling, and optimizing complex systems, he focuses on making AI production-ready and improving developer productivity. His work spans AI strategy, agent platforms, observability, evaluation, governance, security, and enterprise integration. An architect, product leader, and frequent speaker at events such as DeveloperWeek, the Global Big Data Conference, and DEV DAY, Malith shares insights on enterprise AI, software architecture, and emerging technology trends. He holds a PhD in Computer Science from RMIT University, Australia, and has published in respected academic venues including IEEE Transactions on Parallel and Distributed Systems (TPDS) and the Journal of Parallel and Distributed Computing (JPDC).