In a changing digitization landscape, automatic metadata extraction is becoming more important than ever before. At the same time, requirements regarding the extracted metadata are becoming more demanding as well: we are no longer interested in just extracting some data, we want to extract and identify entities. Large leaps have been made on different aspects of metadata extraction, however integrated and effective workflows successfully and efficiently applying metadata extraction to real collections in a market environment are still rare. This paper describes the research, the principles applied and the implementation of just such a workflow.