Code review is the checkpoint where a change is examined — for correctness, security, style, and design — before it becomes part of the codebase. Beyond catching bugs, its underrated value is shared understanding: review spreads knowledge of how a system works across more than one person.
AI changes both sides of review. AI can assist the reviewer, flagging issues and summarizing diffs; but AI-generated changes also make review more essential, because the author themselves may not fully understand code a model wrote for them. A rubber-stamped review of AI output builds no shared understanding at all.
The strongest teams treat review as a comprehension gate: a change should not merge until at least one human genuinely understands it — ideally the person whose name is on it.