I read your blog a lot. Thanks for writing it. But some of the conclusions you're trying to draw in this post feel forced. The "websites -> apps -> bots" deal especially jumps out to me as odd.
many websites, apps, bots conceptually share the same goal: to let users interact with a database. they represent different ways to read and write company data, and as such have different strengths and weaknesses. websites, for instance, are terrific for discovery/browsing activities while chat bots could be terrific for transactional activities (e.g., amazon, cancel my last order).
today, the problem with bots is that natural language AI is extremely difficult, so many criticisms hold true: websites and apps are better than bots for most tasks. today.
but that is similar to someone in 1994 saying that online shopping was worse than offline shopping. yes, it was. but technology improved to where online shopping -- for many cases, but not all -- became better than offline shopping.
the real question: if you believe AI will evolve to where chatting with a company bot is indistinguishable from chatting with the company's top human employee/representative -- e.g., imagine the equivalent of jeff bezos personally handling your amazon queries and transactions -- would a bot become the more natural interface for many activities?