Fix code review after AI tripled the PR volume

Since your team started using coding agents, PRs per week have tripled, and reviewers are drowning. Large AI-written diffs get approved after a quick look, and 3 quiet bugs reached production last month.

One group wants to cap AI use. Another wants to ship faster and trust the tools. Your manager asks you to propose new norms for how the team writes and reviews code.

What do you propose, and what tension does it have to resolve?