CRISP AI — Predicting Reputational Crises Before They Break
An AI service that predicts, identifies, and manages reputational crises at an early stage, reading online and social content alongside live radio and television. Led by Determ, with Bornfight and the Croatian Catholic University.
Reputational crises rarely break without warning. What precedes them is a shift in tone, a rise in emotional intensity, a sharpening of polarisation, a sudden acceleration in how fast content spreads, and unusual levels of engagement. CRISP AI tracks those signals in real time and turns them into an early warning.
The system reads online and social-media content alongside live radio and television broadcasts, and builds four things on top of that stream: an AI-based predictive model, a real-time audiovisual analysis module, a “PR crisis fingerprint” classification framework, and an interface that hands the user an early warning together with recommended response actions.
- DetermProject lead. A Zagreb media-intelligence company whose platform launched as Mediatoolkit in 2014 and now serves communications and PR teams worldwide. Determ analyses the historical crisis data, develops and calibrates the predictive algorithms, builds the real-time radio and television analysis system, and implements the crisis-classification and response-recommendation logic.
- BornfightA digital-product and venture-building company that combines product ideas, specialist knowledge, and technical execution to scale digital businesses. Bornfight translates the research findings into working software: system specifications, interface integration, usability testing, and verification of technical compatibility and performance.
- Croatian Catholic UniversityAcademic and research partner, contributing communication science, sociology, and psychology. The team develops the crisis taxonomy and escalation criteria, defines how the AI model is evaluated, runs focus groups and user validation, and carries the results into conferences and open-access publications.