AI changes the methods available for geopolitical risk assessment and conflict management: qualitative analysis at quantitative scale, real-time consultation with large stakeholder groups in natural language, and the simulation of future interactions to explore possible developments.
The tools below were developed and applied in conflict contexts including Yemen, Libya, and the South Caucasus.
Reading the environment
Analysing news, social media, reports, and other unstructured sources to understand how a situation is developing and being discussed. AI allows this at much larger volume and across several streams at once, while keeping the standards of qualitative analysis.
Draws on semantic search
Monitoring agents read a country's press across its political spectrum - mapping narratives, naming the stakeholders behind them, and flagging what has shifted since yesterday. Every claim comes back traceable to its source. A day of reading, done in minutes.
See it run →Much of public discourse has moved into large messenger groups - hundreds of people in a single thread, hard to follow from outside. The agent sits inside the thread and reports how opinion blocs form, what gets argued, and by whom.
See it run →Listening to people
Reaching beyond the seats at the table to consult communities in large numbers. Current technology gets past the old constraints - access, literacy, information volume - and meets people on their own devices, in natural language.
Draws on stakeholder engagement
Winner of the RTMKM Innovation Prize of the German Economy 2026.
A WhatsApp chatbot runs the consultation in the participant's own language, by voice or text. Hundreds of conversations, analysed for the topics that matter and areas of agreement and disagreement among participants.
See it run →The agent calls participants and conducts the interview as a natural conversation - following up where answers are unclear, and providing information as well as collecting it.
Supporting the process itself
Support for practitioners and diplomats in high-stakes negotiation, in two forms: analysis of what is actually said in the room, and process design informed by how comparable talks have been run before.
Draws on context engineering
The agent works from a session's complete record - every contribution, transcribed - and reports the topics raised, the positions held, and where the room is converging. Each finding points back to the moment it came from.
Advisory agents, each holding one body of practice - stakeholder analysis, historical precedent, inclusion - consulted in preparation or in session.
Projecting forward
Foresight and scenario work ask which paths a situation could plausibly take, and where a strategy stops holding. The aim is not prediction but a structured view of the range of plausible developments.
The agents hold each party's position and play out how a situation could develop - showing which assumptions stand up and which do not.
Six capabilities run through the tools on this page in different combinations - semantic search, sensemaking, context engineering, stakeholder engagement, composition, and translation.
Concept →