Describe when a z-test should be performed as opposed to a t-test

Describe when a z-test should be performed as opposed to a t-test? Which (if any) can we use all the time? Why or why not?

Your manager, who just read an abridged version of a statistics book, wants you to test hypotheses for the difference in two population means. The sample sizes for each are 23. He is adamant that you perform a z-test. What would you tell him? What specific explanation would you give?