Agentic AI moves beyond prompts to systems that plan, call tools, learn from feedback, and close the loop autonomously. This talk unpacks core patterns for Python developers: task decomposition, planning, tool/function calling, state and memory, multi‑agent orchestration, and guardrails. Through pragmatic code examples, it shows how to evolve a simple LLM script into a resilient agent with retries, fallbacks, and observability. Attendees will learn how to wire custom tools and APIs, manage conversational and episodic memory, evaluate and constrain behavior, and productionize agents with tests and telemetry. While cloud‑agnostic at its core, the session also maps these patterns to managed building blocks for deployment, vector search, data access, and event‑driven execution—so teams can ship maintainable copilots, automations, and domain specialists faster and safer.