Share & more
Artificial intelligence is having its messiah moment in the humanitarian sector.
As agencies grapple with funding cuts and layoffs, AI is increasingly framed as the solution that will enable organisations to do more with less – or in some cases, merely survive.
The allure of AI is undeniable, but the reality is more complex. AI may unlock efficiencies in some areas, but its impact is far from straightforward in humanitarian contexts.
In chasing the promise of AI, humanitarians risk sidelining their principles and backtracking on badly needed systemic reforms. But there are ways to harness the potential benefits of AI while minimising harm. Those who fund, plan, and implement humanitarian aid all have a role to play.
Define efficiency: What are we optimising for?
Efficiency, often cited as AI’s strongest selling point, is not a neutral concept: It is shaped by what is prioritised and measured. Does efficiency mean cost reduction? Increased staff productivity? Or is the goal to maximise impact at the community level? These distinctions matter.
In some instances, AI solutions could reduce costs while displacing frontline or country-based staff – for example, in projects that use AI to assess public datasets to generate security assessments more quickly, or in hard-to-reach areas, which could reduce the need for on-the-ground staff and local expertise. While this trade-off might be necessary in some cases, failing to explore where efficiency gains will be made and who ultimately pays could reverse long-standing humanitarian commitments.
Moreover, the assumption that AI inherently reduces costs ignores its hidden expenses. For example, testing and evaluating AI systems – an activity that ensures AI systems continue to perform safely and as intended – require specialised staff and rigorous assurance processes. These carry significant financial costs and time, which are often overlooked or excluded when calculating the costs of AI solutions.
But a failure to properly test and assure AI systems risks increasing harm to the individuals and communities humanitarians are meant to serve. Without deeper analysis, the sector risks pursuing AI for its own sake rather than addressing structural inequities.
A crossroads for reforming the humanitarian system
AI’s rise coincides with a period of profound structural reform. Humanitarian architectures are being re-engineered under crisis conditions, creating both risks and opportunities. On one side, AI could accelerate long-overdue shifts towards decentralisation and shared power. On the other, it could entrench existing hierarchies, reversing gains in localisation.
AI introduces a new power dynamic: Control over data, relationships with AI providers, and access to cloud infrastructure and computational resources are emerging as key determinants of influence.
Many big aid agencies have reaffirmed commitments to localisation while simultaneously championing efficiency via AI. But is the tension between these two agendas surmountable? What would it take to square this circle? If AI adoption is to contribute to positive transformation rather than merely reinforce existing power imbalances, the sector must interrogate whose interests are being served and how decision-making is structured.
Historically, power in humanitarian governance has remained concentrated among donors, institutions, and organisations headquartered in the Global North. Within these organisations, procurement decisions and funding access shape priorities. But AI introduces a new power dynamic: Control over data, relationships with AI providers, and access to cloud infrastructure and computational resources are emerging as key determinants of influence.
What needs to change
To counter the risks of entrenching AI-driven inequities, donors must help smaller organisations – particularly those in the Global South – increase access to AI providers and incentivise partnerships between large and small organisations where skills and expertise on AI are shared.
Humanitarian agencies should be required to transparently disclose their AI usage. And donors should demand that they conduct ethical impact assessments for AI-driven interventions, either through third parties or by using internal experts.
Certification in standards related to the non-technical management and deployment of AI (such as ISO/IEEE 42001:2023, which is designed to help organisations responsibly govern their use of AI) should become mandatory, ensuring organisations deploy AI solutions responsibly. Furthermore, any investments in AI solutions by large humanitarian organisations based in the Global North should be accompanied by efforts to build the capacity of organisations in the Global Majority to deploy AI responsibly. AI must reduce – rather than exacerbate – existing disparities and inequalities.
In parallel, the sector must explore alternative approaches to data collection, funding models, and governance structures that align with a more equitable future for crisis-affected communities. Rewarding agencies who prioritise responsible data stewardship – for example, by involving communities in decisions about which data is collected and how it is managed, or assessing whether data practices reinforce or challenge structural inequalities – could offer pathways towards a fairer system.
AI’s integration into humanitarian work is not simply a technological evolution; it is a systemic re-engineering. Without deliberate scrutiny, AI will reproduce and reinforce the very power structures we should be interrogating. It’s not too late to avoid this fate. But the sector must anticipate the risks and shape these transitions.
