Abstract: Parliamentary data is a publicly available record of South Africa's legislative and political landscape. Bills, committee meetings and ministerial replies are scattered across portals and uploaded inconsistently. Journalists do not have the time to process hours of video, audio, and text data to build stories that inform the public. This paper documents the design, evaluation and implementation of the Parliamentary Intelligence Agent: an AI-powered pipeline that ingests parliamentary data from the Parliamentary Monitoring Group (PMG), cleans and transforms it, clusters the data by topics, classifies the rhetorical intent of parliamentary questions, and produces abstractive summaries that can seed news stories. The workflow comprises four phases: data mining, preprocessing, modelling, and evaluation. The framework utilizes sentence-transformer embeddings with UMAP and HDBSCAN inside a BERTopic pipeline, employs the deberta-v3-xsmall model for NLI-based intent classification, and leverages Google Gemini 2.5 Flash for abstractive summarization.