Amazon may just be the 800lb gorilla that has enough man-hours and computing power to finally integrate all the disparate APIs into a single platform. And their device doesn't cost very much
They're punting though. After a triggering phrase is said it hands off a string to the API associated with the phrase. So, no natural language processing unless the API does it. Users have to be trained to say particular triggering phrases.
The only trigger you have to say is "Alexa", just like Siri or Ok Google. Everything else is processed.