A sector at a turning point

The humanitarian sector entered 2025 under institutional strain marked by the Humanitarian Reset: prioritize life-saving activities, reduce duplication, and strip back non-essential structures. This tension between contraction and transformation defines the environment into which AI is now being introduced. The technology is seen as a potential lever for efficiency gains, better targeting, reduced coordination burdens, and stronger evidence of impact.

However, the study finds that AI adoption is currently being driven less by deliberate organizational strategies and more by the need to do more with fewer resources. Staff are experimenting with AI tools, often ahead of formal institutional guidance, while organizations struggle to translate individual innovation into governed, scalable, and equitable approaches.

Existing responsible AI frameworks are largely comprehensive, but practical implementation support remains limited. Organizations need mechanisms that help them apply guidance, learn from peers, assess risks, and make informed decisions about where and how AI can add value.

In a funding-constrained and legitimacy-sensitive environment, the strategic question is not whether to build new systems, but whether shared infrastructure can simplify the landscape while strengthening responsible practice.

Five findings shape the case for a humanitarian AI lighthouse

1. The challenge is not a lack of guidance, it is making guidance actionable

The sector already has a growing body of responsible AI principles and frameworks. What is missing are the practical tools and support structures that enable organizations to apply them: readiness assessments, implementation resources, use-case evaluation approaches, and pathways for scaling responsible practice.

2. Fragmentation is limiting the sector's ability to learn and scale

Humanitarian organizations are experimenting with AI independently, often finding similar challenges without access to the experiences of others. This limits collective learning and risks widening existing inequalities, as larger organizations with greater resources are better positioned to capture AI's benefits.

3. AI is intensifying existing tensions around equity, accountability, and power

AI does not introduce entirely new challenges but rather amplifies existing ones. Questions around efficiency versus equity, innovation versus accountability, and headquarters versus field decision-making become sharper as organizations adopt new technologies. Responsible AI adoption in practice must therefore address not only technical questions, but also institutional dynamics and humanitarian commitments.

4. The sector needs better curation and accessibility of existing knowledge

Humanitarian AI guidance is increasingly available, but it is often fragmented, difficult to navigate, and designed for specialists rather than practitioners. A curated, practitioner-focused resource base could help organizations identify relevant guidance and translate it into action.

5. Knowledge alone is not enough, communities of practice are essential

Curated resources alone are unlikely to change practice without mechanisms that enable practitioners to interpret, adapt, and apply them in context. Practitioners need structured opportunities to learn from peers facing similar challenges, access to honest accounts of what has and has not worked, practical tools that can be used without specialist expertise, and trusted spaces to discuss emerging practices that formal guidance has yet to address. Content and community are therefore complements rather than alternatives.

The humanitarian AI lighthouse: a proposed model

The report proposes a humanitarian AI lighthouse as a trusted intermediary: not a regulator, technology provider, or replacement for existing initiatives, but a mechanism to connect and strengthen the ecosystem. Its purpose would be to address the operational and coordination gaps that no single organization can solve alone.

The Lighthouse would focus on four core functions:

Building community infrastructure

Creating spaces where humanitarian practitioners can learn from peers, share experiences, and collaborate on common challenges. This could include working groups, peer networks, and facilitated communities organized around AI maturity, organizational role, geography, or operational context.

Supporting practical implementation and responsible AI navigation

Helping organizations move from principles to practice through practical tools, including readiness assessments, use-case evaluation frameworks, implementation templates, and humanitarian-specific approaches to risk assessment and governance.

A key contribution would be a living repository of transparent case studies documenting not only successes, but also costs, challenges, governance approaches, and lessons learned.

Providing market intelligence and technology curation

Helping organizations navigate an increasingly complex technology landscape through vendor-neutral information, tool assessments, pricing insights, and analysis of emerging technologies relevant to humanitarian contexts, including low-connectivity and multilingual environments.

Enabling strategic coordination

Connecting humanitarian organizations, funders, technology providers, and governance actors to reduce duplication, support collaboration, and ensure humanitarian perspectives shape the future development and governance of AI.

A particular priority would be ensuring meaningful participation from local and national organizations, especially those in the Global South, whose operational realities are essential to defining responsible AI practice.

Read the full report to explore the findings, proposed model, and recommendations for building a more responsible humanitarian AI ecosystem.