I’m a humanitarian leader who’s all in on AI. Here’s why

Artificial intelligence is often viewed with deep suspicion in the humanitarian sector due to risks involving bias, hallucinations, and geopolitical misuse. However, as a leader at the International Rescue Committee, I argue that humanitarian organizations must move beyond skepticism to become active, responsible adopters of AI. In an era defined by record-breaking displacement and shrinking aid budgets, we cannot afford to ignore tools that have the potential to scale our impact and save lives in the most challenging environments on Earth.

The current humanitarian landscape is dire, with over 260 million people requiring assistance and 118 million forcibly displaced. Funding has contracted by 25% since 2024, leading to the closure of essential clinics and schools. Amidst this, our initiatives like 'Signpost' have reached 20 million people, while our 'aprendIA' mentor tool is projected to support 6.9 million children by 2028 at a cost of just $0.20 per child. These metrics demonstrate that AI is not merely a theoretical luxury but a practical necessity for operational efficiency.

Historically, humanitarian aid has relied on manual, error-prone processes that struggle to keep pace with the scale of modern crises. From tracking malnourished children to identifying populations missing from census records, the complexity of current disasters demands technological intervention. We have spent years testing AI in refugee camps and conflict zones, learning that while the technology is flawed, its ability to optimize delivery routes and improve data accuracy is unparalleled in the history of modern aid delivery.

Our approach is governed by the principle of 'do no harm,' necessitating a rigorous prototyping process where we aim to break systems before deploying them. For instance, our AI assistant for Signpost underwent six months of testing, improving its safety pass rate from 52% to 77% before human oversight was integrated. We prioritize transparency, publishing our failures to ensure the sector learns collectively. By keeping humans in the loop and ensuring every tool is auditable, we mitigate the risks of bias and data insecurity.

Looking ahead, the humanitarian sector must lead in setting guardrails for AI rather than waiting for others to define the standards. Staying away from AI would mean denying the world's most vulnerable populations access to life-saving innovations. By combining operational discipline with advanced technology, we can ensure that AI serves as a force for good. We must insist on using these tools to solve the world's hardest problems, provided they are deployed with evidence, transparency, and a commitment to human-centric outcomes.