Leashes, Not Guardrails: A Management-Based Approach to Artificial Intelligence Risk Regulation