Abstract. Experimental parameters for an electron microprobe are numerous and their choice influence the quality and accuracy of the analysis results. The expert system intends to optimise the choice of every parameter as well as to automate the stages of a quantitative analysis for all kinds of materials. In short, the expert system aims to master the accuracy of results and to control the time of analysis.The starting point of the expert system is an interactive questionnaire about the sample and about the expectations of the analysis (the accuracy of the result and/or the duration of the analysis). Then, the expert system makes a semi-quantitative analysis on the sample. It gives the sample composition in a first approximation. This is necessary for the expert system to optimise all the parameters for an accurate quantitative analysis. Each parameter is modelled by a specific algorithm. The expert system selects the parameters by minimising, at every stage, the statistical error generated by the algorithm. In this way, it operates in a sufficiently independent way to create a quantitative analysis configuration adapted to an unknown sample.The expert system employs the usual methods of quantification but the results are obtained in half the time. Moreover, the accuracy of the result is guaranted by the statistical error calculation as the expert system is running.