We consider efficient iterative solutions of large-scale electromagnetic problems involving metallic objects. For fast iterative solutions, a multilayer scheme using approximate forms of the multilevel fast multipole algorithm is developed. The approach is based on preconditioning each layer with iterative solutions at a lower layer, while the accuracy is changed from the top layer to the bottom layer. As opposed to the conventionally used algebraic preconditioners, the multilayer scheme: 1) does not require significant setup costs for large problems, and 2) does not require any additional memory. In addition, it can provide faster solutions, especially for large problems. The advantages of multilayer solutions are shown on canonical and complex geometries formulated with the combined field integral equation.