The current condition of soils is a major area of interest due to the lack of certainty in their physicochemical properties, which can guarantee the quality and the production of a specific crop. Additionally, methodologies to improve land management must be implemented in order to address the consequences of many environmental issues. To date, many techniques have been implemented to improve the accuracy—and more recently the speed—of analysis, in order to obtain results while in the field. Among those, Near Infrared (NIR) spectroscopy has been widely used to achieve the objectives mentioned above. Nevertheless, it requires particular knowledge, and the cost might be high for farmers who own the fields and crops. Thus, the present work uses a system that implements capacitance spectroscopy plus artificial intelligence algorithms to estimate the physicochemical variables of soil used to grow sugar cane. The device uses the frequency response of the soil to determine its magnitude and phase values, which are used by artificial intelligence algorithms that are capable of estimating the soil properties. The obtained results show errors below 8% in the estimation of the variables compared to the analysis results of the soil in laboratories. Additionally, it is a portable system, with low cost, that is easy to use and could be implemented to test other types of soils after evaluating the necessary algorithms or proposing alternatives to restore soil properties.