Seperating data transformation functionally from model classes

I'm developing a tool for automated report generation in C++. For that, I read a couple of CSV and text files and store the data in model classes, which only consist of simple getter and setter methods.

On run-time, I have to transform some of the data from the model classes into another representation, which I then have to feed into another system. Currently, the logic for the transformation of the data is done in the model classes itself. I want to separate the logic for the transformation of the data from the model classes so that future changes to the transformation logic don't affect the model classes.

Where would be the best place for that logic? Is there maybe a design pattern for such a case?

Hint: I don't use the MVC pattern.