An agent perceives, decides, and acts, bounded by policy.

Enterprise AI agents are not chatbots. They perceive, decide, and act, with autonomy bounded by policy. This playbook maps the three Microsoft platforms, Copilot Studio, GitHub Copilot, and Azure AI Foundry, the four agent types and the six-layer architecture, the Human-in-the-Loop guardrails that scale, two production case studies, and a twelve-month adoption roadmap, all seen through GitHub Copilot.