Engineering Coordination Cost: How Structural Governance Converts Weak Bundles to Strong

Garicano, Li, and Wu (2026) demonstrate that AI's impact on employment depends on the coordination cost of unbundling human tasks from AI-automatable ones. Their framework treats coordination cost as exogenous. This paper argues that coordination cost is engineerable. We identify three architectural mechanisms — persistent shared context, structural accountability, and cross-task relationship capital — that systematically raise coordination cost, converting weak bundles into strong ones. These mechanisms constitute a distinct governance layer (structural governance) absent from both the AI safety and AI labor literatures. We resolve the tension between AI safety design (which favors statelessness) and economic protection (which requires statefulness) through a layered architecture analogous to the CPU/OS distinction. Companion paper to "Scientist AI Needs a Government" (DOI: 10.5281/zenodo.19417393).