The ongoing digitalization changes the nature of work. Nowadays, even complex tasks can be automated and reliably performed by machines. This new wave of automation has led to an increased interest in predicting the effects of automation on job design. A recent study suggests that around half of today’s jobs could disappear in the coming twenty years. However, these results are heavily debated. Other studies claim that the effect of automation will be much less dramatic. A fundamental issue underlying all these studies is the question of how to categorize tasks. Some authors simply divide tasks into routine and non-routine tasks, others also consider which kind of cognitive abilities are required. Since the predicted effect of automation directly relates to the categories considered, a sound task framework is essential for useful predictions. Recognizing that existing task models are limited in terms of granularity and time, we use a literature study, interviews, and an analysis of historical data to systemically develop a new task framework for predicting the effects of automation. We conduct an evaluation of our framework to demonstrate the generalizability of the framework and compare the framework with existing models.