The Korkine-Zolotareff (KZ) reduced basis is considered as the best reduced basis for use in data decoding in multiple-input multiple-output (MIMO) communication systems. In this paper, we present an improved KZ algorithm for direct application in reducing a complex lattice basis. The enumeration algorithm, which is the dominant contributor to the main cost of the KZ algorithm, is also modified for a complex lattice. We reduce the number of iterations in the KZ algorithm to reduce complexity without degrading performance. In addition, we show that KZ-reduction-aided decoding is capable of achieving full diversity. Simulation results demonstrate the excellent performance and low complexity of the improved algorithm.