Journalists often face the daunting task of manually sifting through vast amounts of documents to uncover newsworthy story ideas. The Djinn platform, or “Data Journalism Interface for Newsgathering and Notifications”, developed by iTromsø, Visito, and IBM, addresses this challenge by leveraging generative AI, supervised machine learning, and rule-based algorithms. Djinn processes municipal documents from Norwegian archives, ranks them by newsworthiness, and generates efficient summaries and visual thumbnails. Its architecture includes a central document ranker alongside local rankers tailored to individual newsroom preferences, enhancing the identification of relevant content. Djinn fosters user trust through explainable AI components and feedback mechanisms. Business results from March-April 2023 and 2024 indicate significant increases in story production and reader engagement for newsrooms using Djinn. This paper discusses Djinn’s design and architecture, the challenges encountered during development—including data availability and privacy laws—and its contributions to computational journalism by illustrating the practical application of AI technologies in investigative journalism.