How-are-humanitarians-using-AI-in-2025
The Humanitarian AI Paradox: Widespread Adoption, Limited Governance
Executive Summary: The Humanitarian AI Paradox
The Humanitarian Leadership Academy (HLA) and Data Friendly Space (DFS) present this joint report on current usage and applications of artificial intelligence (AI) in the humanitarian sector. Drawing on insights from 2,539 survey respondents from across 144 countries and territories, this exploratory research represents the first comprehensive baseline study of AI adoption across the humanitarian sector.
Key statistic:
93% of respondents report using or having used AI tools, with 70% integrating them into daily or weekly workflows. Yet, less than half agree that AI has improved operational efficiency, while only 38% believe it has led to better decision-making.
This pattern, which the researchers term the 'humanitarian AI paradox', describes the disconnect between widespread individual AI adoption and organizational readiness, compounded by mixed individual attitudes about AI's effectiveness.
About the Research
This collaborative research—involving a global survey, key informant interviews, and a rapid literature review—was designed with community participation at its core. The research occurred at an inflection point for the humanitarian sector, with global international humanitarian assistance falling 11% in 2024—the largest cut ever recorded.
Who Participated?
The Five Critical Findings
1. Individual AI adoption outpaces organizational readiness
While humanitarian workers are rapidly taking up AI tools, most organizations remain in early experimentation phases. Only 8% of respondents report AI as widely integrated in their organizations, creating risks around governance and ethical oversight.
Key stat: 70% use AI daily or weekly | Only 22% have formal organizational AI policies
2. Accessible tools, limited specialist expertise
Conversational AI interfaces have lowered barriers to adoption, yet sophisticated AI expertise remains scarce. Only 3.6% of respondents consider themselves to have expert-level AI skills, reflecting the informal, self-directed nature of most AI adoption.
3. Fragmented AI training approaches
64% of respondents report little to no organizational-directed training in AI. 73% identify training as the most crucial organizational support mechanism over the next 12-24 months. Self-directed learning dominates (52%).
4. AI governance vacuum
The prevalence of individual AI usage without corresponding organizational policies creates significant risks around data sovereignty, privacy protection, and alignment with humanitarian principles. 7% work for organizations that explicitly state no intention to adopt AI, while 17% belong to organizations that haven't yet adopted AI but plan to do so.
5. Commercial AI tool dominance
69% of humanitarian workers utilize commercial AI agents (ChatGPT, Claude, Perplexity, Copilot), making these general-purpose tools the dominant form of AI in the sector. Heavy reliance on commercial platforms raises questions about contextual appropriateness and data security.
Mixed Perceptions of AI Effectiveness
Purpose-Built AI Solutions in Action
The report documents four innovative purpose-built AI solutions developed by humanitarian organizations:
🤖 AI Chatbot for Child Caregiver Training (Africa)
An AI chatbot to train caregivers and child/youth professionals. After explaining the anonymous nature of the system, uptake increased by approximately 600%.
📊 AI-Supported Community Livelihood Development (Afghanistan)
An NLP-based data analysis platform for women's economic inclusion programming. Resulted in a 30% increase in completion rates for skills development programs.
🎮 AI-Powered Safety Education for Children (Ukraine)
Interactive educational games teaching safety skills to conflict-affected children. Combines lived experience from conflict zones with AI-generated scenarios.
🔒 Offline AI Assistant for Operational Security (Lebanon)
A fully local AI assistant running entirely offline without cloud dependencies, ensuring complete data privacy and security for humanitarian operations.
Investment and Resource Constraints
Key Takeaways for Evaluators & M&E Professionals
📊 The Governance Gap
Only 22% of organizations have formal AI policies despite 93% individual adoption. Evaluators must assess both organizational readiness and shadow AI risks.
🎯 Training as Priority #1
73% of respondents identify training as the most crucial support mechanism. Evaluators should assess AI competency building as a key performance indicator.
🌍 Global South Leadership
75% of respondents from Global South regions show high engagement and innovation, challenging narratives of technological dependency.
⚖️ Ethics vs. Operational Pressure
Ethical guidelines rank 4th in priorities (43%), behind training (73%), tools (53%), and funding (48%) — reflecting resource constraints.
Future Lines of Inquiry
How can ethics-centered 'ProSocial AI' guide purpose-built humanitarian solutions?
How do specific regional contexts shape AI adoption in humanitarian work?
How can organizations move from pilot projects to scaled AI implementation?
How is trust built in AI tools in relationship-based humanitarian contexts?
Frequently Asked Questions
What is 'shadow AI' and why does it matter?
Shadow AI refers to the unsanctioned use of AI tools by employees without formal organizational approval or oversight. The report finds this is widespread in the humanitarian sector, creating governance risks around data protection, ethics, and compliance with humanitarian principles.
Why do local NGOs have different AI concerns than international organizations?
Local NGOs working closest to affected communities demonstrate heightened concerns about ethical implications and security risks, reflecting their direct accountability to vulnerable populations. They also face dual challenges of limited budgets and lack of technical infrastructure.
How can the sector close the governance gap?
The report calls for coordinated sector-wide approaches including shared policy resources for smaller organizations, expanded communities of practice, greater Global South civil society participation in AI decision-making, and baseline standards that protect both humanitarian workers and affected populations.
What is the most urgent priority identified by the research?
Training and capacity strengthening emerge as the sector's highest priority for collaborative exploration (73%), followed by access to AI tools and platforms (53%), funding and resources (48%), and ethical guidelines (43%).
Why This Matters for the Humanitarian & Evaluation Community
The HLA-DFS report provides the first comprehensive baseline of AI adoption in the humanitarian sector. For evaluators, it offers:
✅ Baseline data on AI adoption patterns across 144 countries
✅ Evidence of the governance gap requiring policy attention
✅ Documented purpose-built AI solutions with measurable impact
✅ A framework for understanding the 'humanitarian AI paradox'
The research concludes that the sector is at a critical juncture where structured dialogue and action plans could help identify pathways for more effective humanitarian response while maintaining core humanitarian principles.
Related Resources
