We present an adaptive compression environment (ACE) that uses both underlying resource performance (network and CPU) predictions and the compressibility of the data stream to perform on-the-fly compression. ACE couples technologies from Computational Grid Computing and dynamic compilation research to predict the efficacy of using compression and to intercept and compress socket-based communication transparently. Our results show that ACE is able to accurately determine when to compress and not to compress. Over all file formats studied, ACE enables significant improvements over solely performing compression or no compression.