This paper addresses the problem of machine selection in the design of flexible manufacturing systems. The proposed heuristic considers the flexibility of machines and the alternate routes for each part to be processed in order to determine the number of machines of each type to purchase and the tasks to be assigned to each of them. The heuristic is divided into two parts. The first part attempts to load and unload machines to improve the annual cost from an initial solution. The second part will attempt to further improve the annual cost by balancing the workload among the machine types. An overview of the heuristic is given and some numerical results obtained for a number of examples are presented. The heuristic generates, in short time, a good feasible solution compared to an optimal one obtained by solving an integer programming model for the same problem. The tests performed show that the proposed heuristic is more performant when used for large size problems.