I wonder what kind of "improvement" you are citing regarding CBC utilities and covariates. Although covariates do seem to make the utilities more distinct across respondent groups based on the dimensions of the segments defined for the covariates, there is a good question about how much is true heterogeneity vs. overfitting. I tend to hope and expect it is mainly true heterogeneity, but I don't doubt that there is some potential overfitting of heterogeneity even along spurious covariate dimensions. Over and over again, there has been no benefit seen in terms of predictive validity of holdouts (whether within-sample or out-of-sample holdouts).
Because ACBC captures much more information at the individual-level (than CBC), and therefore involves less smoothing to the hyperparameters, there has been less interest from the start in seeing covariates built into the HB engine within the ACBC system (they would seem to have less opportunity to do anything). Right now, people can export ACBC results to the .CHO file if they really want to leverage covariates via the CBC/HB utility estimation software.
But, the next version of ACBC to probably be released in Q4 will support covariates within the ACBC software dialogs. Plus, it will be able to do alternative-specific designs, whereas the current version of ACBC cannot.