I wrote about sneaky "pseudo-AI" back in sept, and the problem seems to be getting worse.
If you make an "AI" thing that sometimes uses humans, TELL PEOPLE when they're dealing with a human. Or else your customers will abuse your employees and everyone will be unhappy.
Sometimes people are perfectly happy to deal with an AI. Sometimes they even prefer it (privacy, for example). But you HAVE to make it clear what/who they're dealing with.
You can follow @JanelleCShane.