Discover realistic test coverage targets for AI-generated code. Learn why 80% isn't enough, how to use risk-based tiers, and why mutation testing is essential for quality.