The collection and analysis of data assume a crucial importance in the digital transformation era. Internet of Things (IoT) technologies allow to gather data from heterogeneous sources and make them available for data-driven systems aiming, e.g., monitoring, diagnosis, prediction and optimization. Several applications require that these IoT nodes be located remotely without connection to the electrical grid and being powered by batteries or renewable sources, thus requiring a more efficient management of the energy consumption in their operation. This paper aims to study and develop intelligent IoT nodes that embed Artificial Intelligence techniques to optimize their operation in terms of energy consumption when operating in constrained environments and powered by energy harvesting systems. For this purpose, a Fuzzy Logic system is proposed to determine the optimal operation strategy, considering the node’s current resource demands, the current battery condition and the power charge expectation. The proposed approach was implemented in IoT nodes measuring environmental parameters and placed in a university campus with Wi-Fi coverage. The achieved results show the advantage of adjusting the operation mode taking into consideration the battery level and the weather forecasts to increase the energy efficiency without compromising the IoT nodes’ functionalities and QoS.