Passing Output of one model to another model

I’m doing a XGBoost Regresion prediction model for a chemical process where there are multiple process units and I’m trying to predict the final outcome of product specifications (eg: density of liquid) from the intermediary product specifications (eg: temperature and pressure). I’ve the intermediary prediction model available to me already.

 Raw Material ---> [Model 1] ---->Intermediary Product ----> [Model 2] ----> Final Product


                            Raw Material ---> [Model 3] ----> Final Product                                                                      

The question is if I were straight to predict the final outcome of the product specification using the specifications of raw material bypassing the intermediary product specifications, will it be viable? OR is there a better way to improve the model?

The idea i had in mind was to pass the predicted values of intermediary product from the first model to predict the overall final product specification.