Estimating the task content of work: workforce design for AI-driven human-robot collaboration in intralogistics

This paper addresses the challenge of strategic workforce planning for AI-driven human-robot collaboration (AI-HRC) in intralogistics. We ask two questions: how can task-level full-time equivalent (FTE) estimates be constructed from existing labour statistics, and how can these estimates, combined with AI exposure metrics, inform strategic AI-HRC design and workforce planning? Drawing on U.S. Bureau of Labor Statistics employment data, O*NET occupational profiles, and task-level AI exposure scores, we develop a stochastic task-time framework that decomposes occupations into tasks and models task frequencies as probability vectors on the simplex. A covariance-completion procedure reconstructs task covariance matrices consistent with survey standard errors, enabling the translation of occupational data into task-level and detailed work activity (DWA)-level FTE estimates with uncertainty bounds. Applying the framework to the U.S. intralogistics workforce, we find that approximately 370,000 FTEs (about 17% of workers) are concentrated in the top 15% most AI-exposed DWAs. These results provide task-specific insight into AI-driven automation and support scenario-based workforce planning by linking alternative AI-HRC adoption paths to task-level FTE impacts, uncertainty bands, and upskilling priorities, thereby offering an analytical foundation for resilient, human-centered AI-HRC systems.