Lengthy, but good article on recode with many stats, research and insights about smart speakers and how they are used. It supports my thesis that "voice in, voice out" is just one scenario good for short interactions and information. Voice in, microwaved food out is what Amazon tries with their 60 USD microwave, voice in, video out is what the Echo show tries to do, and voice in, website out is what a Siri search looks like. "Voice in" is probably one of the biggest bets in digital currently, and whoever owns the listening entity, also owns the logic working up any "out" that consumers may adopt. The platform race will be decided between Amazon, Apple, Googe, Alibaba and maybe one or two surprise companies, but anyone who wants to be successful on these platforms needs to understand and serve the logics that are applied behind the voice command. And maybe there will be "voice only" offers that make so much sense with voice, they may make some apps and websites obsolete.
Another source for statistics and numbers - 2017 survey on smart speakers/voice by NPR [PDF]: