The energy consumption used in cooling systems is increasing year by year. In data centers, such consumption can represent more than 50% of the total energy consumption. This work aims to develop an actuator to reduce the energy consumption of air conditioner (AC) systems in a small server room. This work uses machine learning techniques to create an accurate model that predicts the short-term AC energy consumption in the next 30 minutes window based on the inside and outside temperature. Three different actuators were evaluated in this work, the first using the AC unit at 22°C in automatic mode, the second one using fixed rules based on the time and day of the week, and a third using a prediction-based rule system. The case study using rules-based actuators resulted in energy savings compared to the AC automatic mode. In the comparison of the three actuators, the prediction-based actuator model presented the best results, obtaining a reduction of approximately 37.84% in energy consumption compared to the automatic mode.