When Statistics Canada released X-11-ARIMA (Dagum 1980, 1988), the improvements to X-11 (Shiskin, Young, and M usgrave 1967) included an automatic modeling procedure to help users take advantage of new program features. X-12-ARIMA (Findley, Monsell, Bell, Otto, and Chen 1998), the mo st recent program in the X-11 line, retains the X-11-ARIMA automatic modeling procedure, but X-12-ARIMA Version 0.3 includes an additional automatic modeling procedure based on the procedure found in TRAMO (Gómez and Maravall 1997). At the U.S. Census Bureau, we use automatic modeling tools to identify regARIMA models (regression models with ARIMA errors) for forecast extension and estimating regression effects before running the X-11 method of seasonal adjustment. We are concerned that the automatic procedure may identify (1) mixed models that are generally too complicated for concurrent adjustment production at the Census Bureau or (2) models that are susceptible to convergence problems as indicated by coefficient values. We created a program to check automatically-identified models for these concerns and produce X-12-ARIMA input files using simplified versions of the automatically-chosen models. So far, our model-simplification procedure is not completely automated, but we were able to choose simplified models without excessive human intervention.