I think because you get the disjointed Idiocracy Carl's Junior interaction: https://www.youtube.com/watch?v=5d7SaO0JAHk

As it is when I go to some fast food places they greet you in one voice (possibly a central ordering system) and you get a second voice that interrupts (local people I suspect) and takes over. It's weird.

This is a better idea. Put the headset on a delay so AI can inject the requisite politeness. Construction workers don’t have to turn screws by hand so why shouldn’t service workers have a tool that helps them give great customer service?

As a customer, I'd find that actively offensive.

Those pre-ordering recordings asking if I'm using an app are already horrible enough as it is. Offloading basic human politeness to machines would be even worse.

Because the headset doesn't have a loudspeaker mode.

But V2 will! The "AI" will handle the whole customer interaction, with the human doing nothing but carrying it around on his head.

Slightly less-worse take than "isn't it ultimately better if the employees are more polite?"