Architecting agentic AI for the enterprise is challenging because teams must move quickly while designing for scale, governance, security, and flexibility. Successful architectures are modular, deliberate, and adaptable, rather than a monolithic ideal state. Core disciplines such as runtime design, memory, tool use, guardrails, access control, testing, and governance must be developed together rather than in isolation. This report helps architects understand the main components and emerging patterns of agentic AI so they can make better design decisions and avoid costly sprawl and risk.