A recent study from UCL School of Management PhD Candidate and Adjunct Lecturer Shivaang Sharma and Assistant Professor Angela Aristidou, which will be published in the forthcoming 2025 edition of Research in the Sociology of Organisations, examines the ways in which human-AI teams operate in data-sensitive, multi-stakeholder humanitarian crises. 

Drawing on a longitudinal study conducted across diverse sites in Africa, Asia, and the Middle East, the research uncovered surprising coordination patterns. In these regions, teams deployed AI tools—from real-time geospatial mapping and crisis data analytics to remote sensing applications—while working in extremely challenging environments. Their ability to balance a wide range of inputs with rapid, focused decision-making offers vital lessons for managers responsible for leading such hybrid teams.

Cast a Wide Net for Fresh Perspectives

In crisis situations, local insights and cultural sensitivity are paramount. Managers are advised to create open forums that bring together frontline responders, technical experts, and community representatives. This inclusive approach ensures that AI tools are not only technically robust but also tailored to the diverse realities on the ground. For instance, teams in parts of Africa and Asia incorporated localized data to refine algorithmic outputs, resulting in tools that better adapted to regional needs.

Refine and Strategise for Rapid Decision-Making

Once a broad set of insights is collected, clear decision-making becomes critical. Managers should convene a core group—a “strategy cell”—tasked with synthesizing the diverse inputs and rapidly translating them into standardized protocols and tactical decisions. In the Middle East, for example, a designated expert panel was able to streamline feedback from various sources into actionable guidelines that enhanced both responsiveness and accountability during unfolding crises.

Stay Agile with Real-Time Feedback

Continuous adaptation is essential when operating in volatile, high-stakes environments. It is important to establish iterative feedback loops where frontline data flows back to developers and decision-makers on an ongoing basis. This real-time exchange allows teams to fine-tune their AI systems swiftly, ensuring that tools remain effective as conditions change. In several study sites, this agility not only improved operational outcomes but also reinforced trust across the full spectrum of stakeholders.

Acknowledgements

This research was made possible through ongoing collaborations with United Nations Office for the Coordination of Humanitarian Affairs (UN OCHA), UN ReliefWeb teams, the Humanitarian AI community, Data Friendly Space (DFS), iMMAP Inc and Boldcode_. These research-practitioner collaborations are integral to addressing complex societal challenges during a time of escalating global crises.