In this episode of That’s Girl Code, we explore agentic AI systems that go beyond traditional generative models to perceive their environment, reason about goals, take action, and learn over time. We define what an AI agent is, how agentic systems differ from standard large language model applications, and break down the core components that enable autonomous behavior: perception, cognition, action and execution, and learning and adaptation. We then examine key agentic design patterns—including Reflection, ReAct, Planning, Multi-Agent Collaboration, ReWOO (Reasoning Without Observation), and CodeAct—discussing when to use each, their tradeoffs, and the risks of misapplication. We then touch on implementation, highlighting how frameworks such as LangChain, LangGraph, CrewAI, and platforms like AWS Bedrock support orchestration, tooling, memory, and production deployment. Finally, we address the broader implications for software engineers and organizations, including emerging roles, evolving developer workflows, and critical considerations around security, governance, auditability, and ethical responsibility. This episode provides a practical, technical introduction to agentic AI and its impact on the future of software engineering. Github Copilot Agentic demo: https://www.youtube.com/watch?v=onVn-lnHZ9s
