AbstractThis paper presents models for different types of manpower pooling policies. A multi-grade manpower system with mutually exclusive skills is considered. The work load imposed is a random variable characterized by the known joint distribution of the number of jobs to be performed and of the time to do a job. The basic models are developed as a tool for manpower planning in the jobbing workshops of an oil company. The resulting "two-stage programmes under uncertainty" are shown to reduce to mixed-integer linear programmes. The models are then generalized to permit their use in a larger class of manpower planning problems.