The use of spatial augmented reality in manufacturing aims to augment step-by-step work instructions directly into the work environment through in-situ projections. This form of augmented assistance promises to reduce cognitive load, improve learning, and decrease errors, all without calling for any device to be worn, thus increasing work comfort. However, little is known about the interaction with such systems in work scenarios. In this paper, we apply human action recognition to implicitly trigger events based on the performed workflow. We compare this to explicit interaction in an exploratory user study. The results show that actions as implicit triggers could be highly suitable for worker-automation symbiosis.