AI Deployment and Productivity: Why Productivity Diffusion Takes Time

Productivity Research Notes (PRN Series), No. 2612. Generative AI capabilities are advancing rapidly, yet measurable productivity gains may diffuse more slowly across economies. This note argues that the key bottleneck lies not in frontier AI access itself, but in the organizational and deployment capabilities required to integrate AI into enterprise production systems. As firms devote substantial resources to systems integration, workflow redesign, governance adaptation, and deployment-related activities, part of AI's economic gains may initially be absorbed by organizational adjustment costs rather than appearing as broad productivity improvements. Moreover, because many deployment-related activities are service-intensive, firm-specific, and difficult to capture in conventional input measures, measured TFP growth during the AI diffusion process may carry an upward bias that obscures the true scale of underlying organizational transformation.