Signed contracts for the KPN Business Market are spread across several content management system (CMS) and generally not enriched with relevant metadata. In order to help our account managers in quickly and effectively looking for selected contracts we have implemented a solution in which documents from the most important CMS where extracted, converted to text, enriched with metadata and injected in the Elasticsearch engine. Account managers can now search via Kibana relevant documents in a fraction of the time. In this talk we'll describe the challenges we faced and how we (almost) solved them via several python packages (grab, pandas, fuzzywuzzy, etc) and ML algorythms (LSH, CNN deep-learning).