A computer implemented method, system, and/or computer program product generate an optimal training policy and an optimal promotion policy to accommodate an upcoming project. Required new job positions are identified for an upcoming project. A skillset matrix, which describes a current skill level of each worker in a work unit, and a promotion matrix, which describes historical evaluations of how well each worker achieved predefined goals of past projects, are generated. The skillset matrix is utilized to create an optimal training policy that identifies which workers are in need of additional training in order to qualify for one or more of the identified new job positions, and the promotion matrix is utilized create an optimal promotion policy that identifies which workers will be candidates for being promoted into one or more of the identified said new job positions.