How Copilot Answers a Prompt

Written in public

This book is being written in the open. Chapters appear as outlines first and fill in as the controls are verified in a Commercial tenant and a GCC High tenant. Each chapter's "In GCC High" table and the cloud availability matrix in Appendix B show, for every control, whether its status rests on Microsoft's documentation or on an observation, and when.

This chapter sets out how a prompt becomes an answer: the orchestrator, grounding in the Microsoft Graph and the semantic index, web grounding, plugins and agents, and Enterprise Data Protection. It also draws the line that recurs in every later chapter, between the layers Microsoft operates and the controls the customer configures.

Every control in this book answers one of four questions: whose access the feature runs with, what it can read, where its output can go, and who can use it.

In GCC High

Every row below is a documented claim until a tenant check turns it into an observation. The basis column says which, and when.

Status rows for this chapter are added to src/data/copilot/controls.ts as the chapter is written; the table renders here.