We Should All Be Learning From Humanitarian AI
Not for moral goodness, but because that AI is just so powerful
The humanitarian sector is not just full of soft-hearted do-gooders. These people know how to operate under extreme uncertainty and scarcity, and to deliver quite astounding results.
The sector has always been forced to innovate under constraints. Budgets shrink, political winds shift, donor fatigue sets in once or twice every decade; yet crises keep happening. The show must go on.
Anyone who has worked around aid organizations knows the mantra: “do more with less.” We are living through one of those moments now where that mantra is everywhere, in the aftermath of USAID’s retreat and broader fiscal tightening.
But out of scarcity comes invention. Today, humanitarian organizations are experimenting with artificial intelligence in ways that go far beyond hype, and—frankly—are way more developed than what some of their peers in business and finance are doing.
AI assistants are being built to sift through decades of project reports. Machine learning models are being used to anticipate extreme weather events in the Alps. Dialogue systems are emerging to make aid delivery faster, multilingual, and more accountable.
Why should people in finance or business care? Because humanitarian AI is not about moral goodness alone. It is about necessity — creating intelligence under extreme uncertainty. And in a world of increasing climate shocks, systemic risks, and tight capital, that is precisely the challenge financial institutions and corporations face as well.
The Emergence of Humanitarian AI
Humanitarian AI is not a niche sideshow. It’s emerging as a field in its own right, with prototypes and pilots that would look at home inside a hedge fund or multinational corporation.
There’s no such thing as “one” Humanitarian AI. It’s any AI tool applied to, or being used in support of, humanitarian causes.
Some of the most promising use cases are knowledge management and decision support. Aid organizations sit on mountains of reports, evaluations, and project data — most of it unread once published. AI assistants are now being trained to mine this history, surface lessons, and compare strategies across regions and actors. What used to take weeks of manual synthesis can now be done in minutes.
Another stream is predictive modeling. From early-warning systems for floods and glacier collapses in Switzerland, to disease outbreak forecasting in refugee camps, machine learning models are being built to anticipate shocks and allocate resources before disaster strikes.
These are not some academic experiments. These tools are being built and tested in live environments, often under the worst possible conditions, quite literally in war areas and disaster zones where human lives are at stake. And these situations push AI development into places most corporates rarely go (they do, but not to such an extreme extent): multilingual, messy, time-sensitive, and extremely high stakes.
Doing More With Less
The inofficial mantra of humanitarian work is “doing more with less.” People think it’s a recent phenomenon, but it’s been in the humanitarian literature for decades. To give you just a single example: At the moment I’m reading Lynne Twist’s excellent book The Soul of Money, where the author describes how fundraising cycles made work to end world hunger and preserve the rainforests all the more challenging—that was in the 1980s, mind you!
The latest trigger is the USAID retreat, but there have been others: post-financial crisis austerity, post-Cold War drawdowns, and more. Each time, aid budgets shrink while crises expand.
Scarcity is uncomfortable, and it always results in one of two things: more misery, or innovation. The humanitarian folks are (mostly) on the innovative side.
When you can’t hire more staff or extend reporting cycles, you build smarter tools. Humanitarian AI is a direct response to these constraints: automation of reporting, multilingual chat systems that scale across field offices, and predictive models that help prioritize limited interventions.
The lesson here is that efficiency isn’t a luxury; it’s survival. In humanitarian contexts, every dollar not wasted means a family gets food, or a flood response arrives in time. The “efficiency premium” is existential, both for humanitarian agencies, but even more so for the people they serve.
Finance and business should take note of this. While the threats ahead are not as extreme (and never have been) as in humanitarian work, we’re definitely seeing some crises on the horizon. Capital isn’t as cheap as it once was, politics and the economy are going through their usual schtick and not getting better; and then there’s all the social issues in ever-fragmenting societies and, well, climate change with all that entails…
And then there’s investors who nevertheless demand both returns and resilience. Good luck with all that! The good news in facing all this is that Humanitarian AI has already gotten busy with showing us some of the answers to all this.
Takeaways for Finance & Business
So what does all this mean for financial institutions and corporations? Three things stand out.
First, learn to act with incomplete information. Real data is messy even in the most cushy environment; I know this first-hand from particle physics (data doesn’t really get richer and cleaner than it that part of academia). In academia, you can just wait six months until the particle detector delivers fresh data. In humanitarian contexts, people would die if you did that! Waiting for perfect data is a luxury that just doesn’t exist. Aid workers have to make calls with only fragments of knowledge, and AI helps structure those fragments into something actionable. Finance and corporates increasingly face the same: climate stress tests, geopolitical shocks, fragile supply chains, and many more vagaries. Decisions can’t wait for a full dataset.
Second, build transparency into AI. Humanitarian AI assistants are designed to cite sources, show confidence levels, and explain their reasoning — because real lives depend on trust. Investors and clients will expect the same from financial AI systems. Black-box models won’t cut it when portfolios are on the line. Advanced techniques, both on the qualitative and the quantitative side are needed. (My physics background does help me and my team make a meaningful contribution here.)
Third, efficiency is not optional. In a tightening capital environment, the winners will be those who extract the most value from scarce resources. Humanitarian AI shows how necessity can drive precisely this kind of efficiency: faster reporting, clearer communication, sharper prioritization.
In short: humanitarian AI isn’t cute charity tech. It’s a living example of how organizations under pressure reinvent decision-making — a lesson finance and corporates do well to absorb.
Extreme Weather and Systemic Risk
One domain where the overlap is particularly clear is climate risk. Humanitarian AI models are being built to anticipate floods, glacier collapses, and disease outbreaks. Insurers and asset managers are doing the same — only their goal is balance sheets rather than lives. (Okay, that’s not entirely true—insurers in particular very much do protect and save lives, and asset managers do so indirectly, too. But the point is, they’re not the orgs at the frontlines.)
During my time at AXA Climate, I saw firsthand how closely these two worlds align. Early-warning systems for extreme weather are just as relevant for insurers as for humanitarian logisticians. Proper modeling of complex weather events (that was one of my responsibilities at the time) fuel parametric insurance tools that trigger payouts after disasters, which stabilize both households and corporate supply chains. And predictive models that prioritize interventions in fragile areas can just as easily be applied to portfolios exposed to climate-sensitive industries.
The truth is that extreme weather is a systemic risk. It doesn’t care whether you are a villager in Senegal, a CFO in Mumbai, or an insurer in Zurich. Humanitarian AI and financial AI are both grappling with the same problem: how to prepare for shocks that are certain to come, but uncertain in timing and form.
Here’s the opportunity: instead of reinventing the wheel separately, finance and humanitarian actors can learn from each other’s approaches — and build shared resilience.
The Bottom Line: Humanitarian AI Is for Us All
Humanitarian AI is a testbed for the future of resilience. If algorithms can help aid agencies allocate scarce resources, make sense of messy multilingual data, and act under severe time pressure, then the same principles can help financial analysts, insurers, and corporates manage systemic risk.
The value lies less in the technology itsel,f and more in the mindset that it forces. Working under extreme pressure (human lives quite literally depend on this work), with incomplete data, explaining decisions transparently, and prioritizing what actually causes outcomes rather than what merely correlates with them.
I’ve seen the overlap firsthand in my time at Axa Climate, where models built for insurers and models built for disaster relief often tackled the same underlying question: how do you prepare for the unpredictable? The answer was never to eliminate uncertainty, but to manage it with clarity, speed, and trust.
So when we talk about humanitarian AI, we’re not just talking about saving lives in faraway places. We’re talking about the future of decision-making everywhere.
Finance, corporates, and aid organizations share the same problem: shocks are coming faster, resources are tighter, and trust is harder to win. We’re aware of this, and are working on some collaborations (to unfold more in the coming months). I’m convinced that in times like these, the smartest move now is to learn from each other.
Reads of The Week
This story from Zimbabwe offers powerful inspiration for inclusion and equity across Africa. The Ezekiel Guti Jnr Legacy School was born out of one man’s unwavering dream to create a safe, dignified space for children with disabilities—many of whom face stigma and abuse in their communities. Beautifully penned by Humanitarian Eye, it’s an inspiration for all of us to dream with our hearts and then collaborate and partner up to make that dream real.
The crisis in the humanitarian sector goes beyond USAID. With UN funding at a record low—only 19% of what’s needed—its ability to respond to crises in places like Gaza, Somalia, and Mozambique is collapsing. As Kevin McSpadden puts very clearly, it’s more urgent now than ever to build more resilient, locally grounded aid systems while also pressing for global solidarity in the face of rising need and shrinking support.
In this reflective essay-style post, Rt Hon Steve Baker FRSA introduces two classic libertarian texts—Frédéric Bastiat’s The Law and Isabel Paterson’s The Humanitarian with the Guillotine—as a lens for critiquing state intervention and the unintended harms of benevolent policies. It’s a great piece to remind you of the dark side of humanitarianism; that bad things are often done by good (if ignorant) people.
Thanks for the shout-out!