Present challenges of small batch production are represented by the need to improve time-to-market and the reduction of costs. A promising approach to take up these challenges is the use of highly iterative development processes such as Scrum known from software development. A transfer of these principles to process planning enables the prediction of producibility of customer orders by iteratively learning from manufacturing data of similar jobs from the past. Based on the required data structures described in this paper, work plans for new orders can be generated automatically. The potential of the approach is validated by an industrial example.