From AI-Infused to Agentic: Building an Enterprise Agentic Framework

Over the past year, the rise of agentic architectures has pushed the boundaries of how we design AI-infused applications. What began as simple wrappers around large language models has evolved into more autonomous, stateful, and interactive systems, agents capable of planning, invoking tools, and collaborating.

In this talk, we’ll share our journey from building traditional AI-enhanced services to designing a full-fledged agentic framework. Along the way, we faced new design challenges: How should we model local and long-term memory? How do we represent multi-turn exchanges while preserving context and state? What protocols are needed for interoperability, such as the Model Context Protocol (MCP) or Agent-to-Agent communication, and how do they fit together? What about observability, security, auditing, and developer experience?

We’ll walk through the core concepts and patterns we explored, including function calling, memory management, supervision, planning, and workflow orchestration. We’ll discuss the architectural trade-offs we encountered and the lessons we learned along the way. Whether you’re building your own agent framework or just curious about what makes agents different, this talk will offer practical insights from the trenches.