Bat species are an integral part of our ecosystem and their monitoring can provide important insights into conservation and tracking viruses like Covid-19. Given the difficulty and high cost of manually monitoring bats in their natural habitats, this paper proposes an Artificially Intelligent Internet of Things (AIoT) system that uses audio-based Convolutional Neural Network (CNN) to monitor bat species using their echolocation calls. The system uses Long Range Wide Area Network (LoRaWAN) to send the classified species to an application server in real-time. The paper compared the performance of three different edge devices, Raspberry Pi Model (RPI) 3B+ (RPi), NVIDIA Jetson Nano, and Google Coral and two deep learning frameworks (TensorFlow Lite and TensorRT). Although all edge devices were able to do the real-time inference (<; 0.5 seconds/inference for a 3-second audio segment), Google Coral appears to be the best choice because it was the fastest (0.3917 seconds/audio segment) and required the least resources (maximum %CPU Utilization = 29.2%). However, if cost was a concern then even the RPI was more than adequate for the task.