This article will be of those interested in tracking the progress toward robot tutors (a.k.a. AI assistants). The first part of this article offers a taxonomy of five levels of personal assistants. To date, our best is level three: "those that can handle naturally-sounding, multi-turn interactions, are hard to build." The second part of the article describes in some detail how to build a level three assistant, outlines some of the challenges (entity disambiguation), and next steps. By the time I'm at the point where I can think of incorporating AI assistants into a personal learning environment, we'll probably be at level four.