Precision agriculture is aimed at field management considering its spatio-temporal variability. Its widespread use has been made possible with the development of tools for data collection and georeferencing of productivity, soil properties and others. The large amounts of data generated require the use of information technology resources for processing, allowing better definition of management zones. The correct selection of parameters is a complex task due to the large number of interrelated parameters, resulting in a nonlinear problem which, associated to the inherent problems in data collection make it appropriate to use statistics and computational intelligence techniques in their approach. Our aim was to compare some algorithms for delineating management zones. Fuzzy c-means, expectation maximization, X-means and self-organizing map were used. Georeferenced point measurements of physical and chemical properties were obtained from a 134.2 ha field. The properties used were pH, Ca, Mg, K, SB, CEC, P, C, OM, V, clay, silt and sand. The data were interpolated to a 10m x 10m grid. The software platforms used for delineating the management zone were Weka, Management Zone Analyst and Matlab, which together provide different algorithms. They were applied to a set of soil attributes and the result obtained shows differences between the techniques used. Some algorithms, such as expectation maximization, provided an excessive number of management zones when the number was not defined by the user. In addition, the algorithms delineated different number of management zones, depending on the soil properties used and on the parameters or the structure of the method. Initial results show differences between maps generated by linear and nonlinear methods. Moreover, some of them require the user to choose parameters or/and structures, imposing a complex procedure to the end user. Final analyses will determine more differences between used methods.