The Internet of Things (IoT) is a massively growing field with billions of devices in the field serving a multitude of purposes. Traditionally, IoT architectures consist of “edge” sensor nodes which are used purely for data collection, actuators for intermediary connectivity, gateways for transmitting data, and cloud servers for processing. However, as application requirements change, IoT architectures must evolve. In the next generation of IoT, traditional architectures may not hold due to power limitations, privacy concerns, or network reliability in certain environments. In particular, Artificial Intelligence (AI) and Machine Learning (ML) applications – especially image/video processing – have become increasingly widespread, and are famously data and compute intensive applications. In this paper, we give an overview and describe the shortcomings of traditional IoT architectures, and outline key scenarios where local processing may be preferred over transferring data to the cloud. We then describe and evaluate metrics which designers can use to assess the needs of their IoT platform. This framework will be beneficial to IoT device manufacturers and network architects for improving usability and reliability, while protecting privacy in the next generation of IoT.