I think there is a huge discoverability problem created by the limitations of these devices. Because we don't believe it can understand anything we say, we instead limit our vocabulary and interactions to what we are confident they will understand. It's the opposite of AI in that this is the machine training us how to communicate with it. So we sit there barking out commands because we are unsure of what it can handle, creating a giant hunt-and-peck problem.
This, to me, is likely why Amazon sends me emails saying "What can Alexa do now?"
This. I had to disable google's "assistant" notifications because it kept pestering me about things I had no interest in. They always seem to have an "I see you're trying to write a letter. Do want some help?" clippyesque feel to them. When I activate a system is probably the only time it's ok to add a guiding interaction (like do you want me to make this default?). I really wish I could tell the voice activated things, "Ok, from now on when I say to play 'rockin beats' I'm talking about my Pandora channel, don't make me say 'play rockin beats on pandora' every time." I do almost the same things every day with Alexa, but it still messes up simple context. For instance: I will never ask for alexa for news while THE ALARM IS GOING OFF. If it "hears" news while the alarm is going off then it heard wrong. I said SNOOZE.
That’s exactly the direction Apple is taking with their Siri Shortcuts, it can tell when you’re performing actions and then later suggest them in similar circumstances.
Anecdote: I've discovered features with Alexa on accident before. I was watching Jeopardy at my parents house and someone yelled that Jeopardy was on, and Alexa was like "let's play Jeopardy!" and booted up a game for us. I would have had no idea that Alexa could do that if it weren't for the fact that it's.. uh.. constantly listening to everything that goes on in your house.
This, to me, is likely why Amazon sends me emails saying "What can Alexa do now?"