This paper describes my PhD project and its advancement in the first months of its first year. The PhD aims to study holistic methods for building, populating, and exploiting warehouses of heterogeneous content. Each warehouse is characterized by a specification of the types of content we search for; a set of websites in which to search for the content; a set of dedicated methods to analyze and understand the content, including to establish or find links that connect different pieces of content. AI and uncertainty are naturally involved in these steps. We present the overall thesis aims, as well as encouraging preliminary results for one use case: the acquisition of statistical data resources from French government websites, leveraging reinforcement learning.