Some 50,000 news items on commodities are published every day – more than any analyst can read, and general-purpose language models read them wrong: a fatal accident on an oil rig scores as negative news, while the market reads it as lost supply and a rising price. NewsQuant (formerly NewsForce) is a research project at FH Graubünden training commodity-specific models that tell those two things apart. We designed the platform that puts them to work.
An agentic platform instead of a dashboard full of filters. Analysts ask in plain language, the agent asks back for whatever it still needs – asset, timeframe, type of visualisation – and answers by building the view: news ranked by their correlation to the Brent price, a forecast with its confidence band, market scenarios weighted by probability and impact. Every answer arrives as a card you can keep, compare and export. Concept, conversation design and agentic UI, grounded in expert interviews and validated in usability testing. A Master UX Design project at FH Graubünden, 2025.











