The study aims to explore the feasibility of adopting for inflation forecasting a sophisticated expert system normally used in routine outlier detection and deseasonalization of time series. Known as TRAMO/SEATS expert system, this twin program is a fully automatic procedure that extracts the trend-cycle, seasonal, irregular and certain transitory components of high frequency time series via the so-called ARIMA-model-based method. The results of the study reveal the feasibility of the use of the technique for routine inflation forecasting. The automatic model building capability of TRAMO/SEATS is exploited to arrive at an ex-ante model that has the ability to generate optimal forecasts. The results show the ability of the final model to forecast inflation with remarkable accuracy.