Presumably they will use approximations like factor analysis that are already commonly used in data science.