How do you address lack of context when generating email responses using large language models

Question

Can i know How do you address lack of context when generating email responses using large language models?