Automation in smaller enterprises requires methods for robot programming usable by non-experts. In this paper, we propose an approach based on finite state automata. The basic principle is kinesthetically guiding the robot through tasks and programming state transitions to represent absolute movement, relative movement (to the current pose or to that of an object), or interaction with saved pose variables. The automata structure allows us to represent control flow (loops and branching according to conditions in the perceived scene). We forgo a visual user interface to avoid switching between input devices. This can save time, aid concentration, and reduce the amount of hardware in the workspace. In our approach, the user executes specific key movements while guiding to trigger operations. We assume that users can program small, well-known tasks only seeing the current scene. The approach was evaluated in a user study. The results show that it can be used effectively, but could be improved upon in terms of intuitiveness.