Generative vs Discriminative

Let’s say you are part of the Gmail team at Google.

Your manager asks your team to develop a classifier to tag emails as spam or not spam.

One of your colleagues suggests using a Naive Bayes classifier. But another colleague disagrees because Naive Bayes is a generative model, and they say it would be better to use a discriminative model like Logistic Regression.

How would you explain to your manager the difference between a generative and discriminative model? And which one of your co-workers is correct?