A granular activated carbon (GAG) column is an effective treatment technology for the removal of lead. However, this technology requires time-consuming and expensive bench- and pilot-scale studies to design a full-scale system. A virtual adsorber system (VAS) based on artificial neural network technology was developed from 67 bench-scale experiments as a new tool to optimize the GAC process. In addition, VAS can be used to design a full-scale adsorber system by eliminating the need for further lengthy and costly experiments. Data obtained from the VAS indicated that decreasing the influent lead concentration from 50 to 1 ppm increased the number of bed volumes (BVs) of wastewater treated at breakthrough from 30 to 950 BVs and exhaustion from 200 to 1650 BVs, while the surface loading decreased from 17 to 1.8 g Pb/g carbon. In addition, increasing the empty bed contact time from 1.85 to 12.75 minutes for each influent lead concentration increased the bed volumes of wastewater treated at breakthrough, while the bed volumes at exhaustion decreased and the surface loading slightly changed for the lower Pb concentration (1 and 10 ppm of Pb). Five sets of training data were selected to test the VAS. It was found that the VAS could predict the bed volumes at breakthrough and exhaustion, and surface loading with an accuracy of 97%. The average coefficients of correlation, R, between actual and virtual bed volume measurements at breakthrough and exhaustion and for surface loading were 0.988, 0.980, and 0.988, respectively, for the verification data, while they were 0.996, 0.994, and 0.996 for the training data. The high values of the correlation coefficients demonstrated the high performance of the VAS for the removal of lead.