Environmental monitoring of coastal marine ecosystems is a routine activity in many countries. In Italy several institutions have been involved in this task, both at regional and at national level, and a very large amount of data is now available. Although the most recent Directives issued by the European Union (Water Framework Directive and Marine Strategy Directive) stress the role of the biotic components in the evaluation of marine ecosystems, ancillary information about nutrients, salinity, turbidity and chlorophyll concentration plays an important role in routine monitoring activities and it can be regarded as a proxy for more complex ecological properties. In other words, these easy to measure variables can be used for evaluating the potential ecological status of coastal marine waters. The CAM expert system (this acronym stands for 'Classificazione Acque Marine', i.e. 'Marine Waters Classification' in Italian) was developed and thoroughly tested in order to easily provide such an evaluation. The first version of the CAM expert system was based on the recognition of 6 basic typologies of coastal waters, which had been defined by means of a k-means non-hierarchical clustering. These basic typologies were analyzed and ranked according to their ecological properties in order to define a two-fold ranking scheme that simultaneously accounted for trophic state (potential productivity) and phytoplankton biomass (actual productivity). The final evaluation was then corrected on the basis of several empirical rules aimed at detecting anomalous combinations of values in data records (e.g. high turbidity in high salinity waters is often due to heavy sea conditions before sampling). The CAM expert system is now in use at the Italian Ministry of the Environment for routine classification of data records collected at 81 coastal monitoring sites (http://www.minambiente.it/index.php?id_sezione=1110, in Italian), but it is also available for downloading as a stand-alone, user friendly application. Further developments of the first version of the CAM expert system, based on other Machine Learning techniques (Self-Organizing Maps, Support Vector Machines, fuzzy c-means classification), will be also presented.