2.3.3. Code Review and Coding Assistance

💡 First Principle: Copilot code review reads a change in the context of the whole project, not line by line in isolation, which is why it can catch issues a linter can't — and why it pairs naturally with the coding agent that can then fix them.

Copilot provides two complementary things here: everyday coding assistance (the suggestions, Chat, and edits you've already met) and Copilot code review, an AI reviewer for pull requests. Since moving to an agentic architecture (March 2026), code review gathers full project context before analyzing a PR, understands how the change relates to the broader codebase, and can hand identified issues to the coding agent to generate fix PRs — a closed loop of find-and-fix with you approving the result.

Organizations can extend code review coverage to all pull requests on GitHub.com, including PRs from authors without a Copilot seat (an admin-enabled policy, with usage billed to the org). This gives teams complete review coverage without buying a seat for every occasional contributor.

✅ Best Practice: Treat Copilot code review as a fast, tireless first-pass reviewer that surfaces context-aware issues — then apply human judgment. It augments reviewers; it doesn't replace accountable human approval.

⚠️ Exam Trap: Copilot code review is not a guarantee of correctness or security; it's an assistant whose findings still need human validation. And enabling it for non-licensed users' PRs is an explicit admin policy, not the default.

Reflection Question: How does an agentic, project-aware code review differ from line-by-line linting, and what makes it able to participate in a fix-it loop with the coding agent?