The idea of employing deep autoencoders (AEs) has been recently proposed to capture the end-to-end performance in the physical layer of communication systems. However, most of the current methods for applying AEs are developed based on the assumption that there is an explicit channel model for training that matches the actual channel model in the online transmission. Since the actual channel varies over time, this imposes a major limitation on employing AE-based systems. In this paper, without relying on an explicit channel model, we propose an adaptive scheme to increase the reliability of an AE-based communication system over different channel conditions. More precisely, we divide the interval of random channel coefficients into n sub-intervals. Subsequently, in the offline training phase, we employ an AE bank consisting of n pairs of encoder and decoder and perform training over the sub-intervals. Then, in the online transmission phase, based on the actual channel conditions, the optimal pair of encoder and decoder is selected for data transmission in terms of satisfying an average block error rate (BLER) constraint imposed on the system. To monitor actual channel conditions for adopting the adaptive scheme, we assume a realistic scenario where the instantaneous channel gain is not known to Tx/Rx and it is blindly estimated at the RX, i.e., without using any pilot symbols. Our simulation results confirms the superiority of the proposed adaptive scheme over a non-adaptive scenario in terms of average power consumption. For instance, when the target average BLER is equal to 10−4, our proposed algorithm with n = 5 can achieve a performance gain over 1.2 dB compared with a non-adaptive scheme.