With the collaboration of Universidad Pompeu Fabra (Barcelona), Eticas carried out an audit of the natural language processing (NLP) system from the Social Services area from the Barcelona City Council. The NLP system was developed in order to categorize the transcribed verbal content of interview performed to citizens by social workers. The purpose of the data treatment was to classify into defined groups the nature of the different citizen’s demands in order to streamline the resource assignation to supply for the population’s needs by reducing the human cost necessary for this allocation. The contribution of this collaborative study between Eticas and the University consisted, on the one hand, in evaluating the accuracy of the deep learning model accorded to the demands, problems and resources’ classification from the interviews collected by the social workers. On the other hand, the audit intended to elucidate if the system could have a negative differentiated impact in certain socially disadvantaged groups, fruit of an incorrect or biased association in certain types of demands, problems and resources.