Which you choose will depend on the context.
• Is there only one consumer?
• Does it make sense to wait for a result?
• Which way do I want my dependencies to point?

I guess you can see it that way.

The point is:
In order to specify your intent to the back-end, you must specify *what* you're doing.
Otherwise, everything's a generic update, and you can't chaing logic, policies or events to it.

In a nutshell: Instead of designing the UI to send a whole form with the latest data, you design it so specific parts are independently sent.

Why?
Because you can design better logic if you do ChangeUserEmail or DisableUser, instead of just UpdateUser with the same information changed.