A fast novel potentiometric titration procedure is proposed, in which the detection system is formed by a potentiometric electronic tongue. The titration consists in a reduced number of fixed titrant additions to the sample and the recording of the potentials of an array of Ion Selective Electrodes. The obtained data matrix is entered to an Artificial Neural Network response model, previously trained to furnish concentrations of a multicomponent mixture. The principle is demonstrated with automated EDTA titration of mixtures of Ca 2þ and Mg 2þ at fixed pH 8.5. In this case, five 2 ml fixed volume additions plus the readings of four sensors were adequate for the resolution of the mixture. In the conditions used, titration was feasible up to 3.3 mM total ion concentration, the precision was estimated as 3.20% RSD for Ca 2þ and 2.76% RSD for Mg 2þ (n ¼ 5), and the detection limits 0.16 mM Ca 2þ and 0.26 mM Mg 2þ . The procedure was applied to mineral waters and compared with reference methods (correlation ¼ 0.92 for Ca 2þ and correlation ¼ 0.89 for Mg 2þ , n ¼ 14).