Why are Copilot prompt-a-thons not driving real productivity? Because the Prompting Plateau is an incentive problem, not a literacy one. Employees won't use AI for deep analysis when their performance review still rewards emails sent and decks delivered. The fix is rewiring incentives, training depth on one real workflow, and replacing active-user metrics with audited financial impact.