It's a problem that will be solved better and better as technology improves. GPT-3 is holding a lot of promise in taking input from one domain (language) and generating an outcome in a different domain (commands).

Not sure how bad/good Alexa is, but everyone (myself included) in my household are talking to Google and Siri all the time to do tasks that would have taken multiple clicks on a screen, and a bunch of typing. A huge reduction in friction. The state of the art isn't perfect, but as it improves, it's undoubtedly taking UX in a direction that is popular and desirable to consumers.