The norm architects: How AI delegation norms form before anyone designs them

Executive summary Organisations introducing generative AI rarely design how they will delegate to it; they discover the pattern after the fact. Convenience choices accumulate, spread by observing leaders, and harden through tipping-point dynamics into implicit norms about which reasoning tasks humans retain or cede. By the time governance frameworks appear, a default norm already dominates. Drawing on research in social norms, cognitive automation, and AI-mediated decision-making, this article explains how AI delegation norms emerge, stabilise, and resist revision. It identifies the pre-design phase as the lowest-cost moment for managerial intervention and shows why three common frames – productivity, capability attrition, and agency – each target the wrong unit of analysis to shape a norm. It proposes three governance practices and a brief Delegation Audit to support proactive norm design.