I think it would be fair to doubt your point, could you share which earbuds you were using and how you were using them? I think the disagreement here will also stem from the fact that IEMs + playing music is pretty great “active noise cancellation” in itself.
When I listen to a podcast on my IEMs I hear quite a lot of the outside world, when I do the same with ANC headphones on I hear much less.
A lot of the form factor is already mostly available in smart watches. They have to, at the bare minimum, conver the bsse functionality of those before moving onto real time ai interaction that is never real time and is hardly a proper interaction.
Progressive enhancement would be great here, smart watch in a pin form factor but with AI powered features when they make sense. Maybe some kind of super fine tuned orchestrator that know when to pass onto siri/assisstant vs. some cloud model (setting a timer requires simple parsing but a complex philosophical question can be offloaded to AI)