Deeplomacy: innovation doesn’t always start in a garage

In early 2022, the world was shifting fast. Information was abundant, yet insight was scarce. Governments and institutions struggled to understand the potential geopolitical challenges, ranging from imminent conflicts to rising waves of violence. The signs of a Russian attack on Ukraine were visible and widely discussed, yet no country took meaningful steps to de-escalate the situation. Despite the abundance of data, few were able to focus on the right signals. That realization marked the beginning of what would later become Deeplomacy: the idea that data and information, if properly analysed, could help us anticipate global challenges before they unfold.

During those tense weeks, Paola Pisano, former Minister of Innovation, was working within the government. Witnessing how difficult it was for institutions to interpret complex global dynamics, she wondered if artificial intelligence could help organisations detect early signs of instability. Through her work as a professor at the University of Turin, she connected with Luca Macis and Marco Tagliapietra, two master's students in Stochastics and Data Science, who were fascinated by complexity and driven by a passion for AI systems.

They began building a model to explore whether early signals of conflict could be detected from open data. At first, it was just an academic exercise — an idea to challenge what AI could reveal from the noise of world events. But the result surprised even them: patterns suggesting an escalation in Ukraine were visible more than two months before the invasion.

That discovery changed everything. It wasn't just a statistical anomaly, but a glimpse of how data, when collected and analysed properly and in vast quantities, could provide foresight to drive action. The team published a paper and soon found themselves in conversation with analysts and policymakers, including experts from the Crisis Unit of the Italian Ministry of Foreign Affairs.

Encouraged by that success, the group expanded its vision. They integrated new datasets, added layers of semantic analysis, and built a pipeline capable of collecting, tracking, and classifying hundreds of sources. This multitude of data supports not only the emergence of a diplomatic network, but also the monitoring and prediction of its evolution, trends, and sudden shifts — through a reliable early-warning system, tested with real-world feedback and refined through collaboration with domain experts, journalists, and institutions.

By late 2023, the team had taken a major step forward. Working side-by-side with the Data Science team at Intesa Sanpaolo, we refined our model to make it not only more reliable, but finally understandable to the people who rely on it. That collaboration turned into a paper that went on to win the Public Award at the 2nd World XAI Conference.

In May 2025, the team officially launched Deeplomacy — an academic spin-off and startup dedicated to transforming complex global data into geopolitical intelligence. Its mission is to help governments, NGOs, and businesses anticipate geopolitical challenges and instability, and to react proactively and strategically.

By combining scientific rigour, machine learning, and domain expertise, Deeplomacy turns vast data flows into actionable insights — empowering analysts and decision-makers to act earlier, with confidence and context.