Change in health systems starts at the top: what we took away from UN General Assembly week

By Ram Fish, 19Labs Founder & CEO

Every September, New York becomes the easiest place in the world to meet a minister of health. For one week, the leaders who decide how health systems run, what gets funded, and which technology reaches patients are within a few blocks of each other. For a company that partners with governments, there is no more valuable week of the year.

This year I spent the week with ministers and health leaders from countries that are moving fast. Midtown traffic during UNGA is its own lesson in gridlock. Everything moves faster than the cars, and the quickest way from one meeting to the next is to walk. Fortunately, the conversations moved faster than the traffic. Their health systems look very different, but the leaders I find most interesting share one trait: they treat new technology as a national decision, not a side project.

Why it has to come from the top

The pattern repeats everywhere. Most promising technologies never even reach a pilot. There are too many obstacles, and too many consensus-driven decisions that get stuck before anything moves. And even when there is a pilot and it works, it often stalls there. Not because the tool failed, but because nothing around it changed: procurement, budget lines, staffing, data rules, and who owns it once the grant ends.

None of those are technology problems. They are leadership decisions. A district can run a pilot. Only a ministry can turn it into infrastructure.

When a minister of health decides that rural communities will get specialist access, connected diagnostics, and medicines delivered by drone, the system lines up behind that decision. Workforce plans adjust. Connectivity gets prioritized. Regulators engage early instead of late. That is the moment technology stops being an experiment and starts serving people.

We have seen this in Guyana and Mexico, where government partners made GALE part of how care is delivered, not a side project. In Guyana, our partnership with the Ministry of Health has grown GALE from individual sites into a national network reaching communities across the country. In Oaxaca, Mexico, the state health network went from fixed sites to mobile units, and now to GALE | Toci, our AI health agent that guides people to the right care.

AI was in every conversation

AI came up in nearly every meeting. That is no surprise. What stood out is the distance between the global conversation about AI and the reality of a rural health post hours from the nearest doctor.

I was also in the room when Tony Blair and Bill Gates sat down at the Tony Blair Institute for Global Change. The conversation centered on a real risk: that frontier AI models could be abused. Gates rightly emphasized the need for monitoring. What was not discussed is how governments can actually use and deploy AI for their people. That gap matters. When the public conversation is mostly about fear, governments hesitate, and hesitation means falling behind instead of using AI to improve people's lives. Both conversations are needed. Only one of them is happening.

A more useful frame came from Ezra Klein's latest episode, "Intelligence Isn't Power," with Princeton's Arvind Narayanan, co-author of "AI as Normal Technology." Intelligence and power are separate things. AI looks enough like past transformative technologies that we already have a roadmap for adopting it, one domain at a time, inside real institutions. That is the kind of guidance governments can act on: pick a domain like rural health, build the system around it, and deploy.

Our view is practical. AI in health does not start with a model. It starts with a system that can use it: a local health worker with a connected diagnostic device, a telemedicine link to a specialist, and a supply chain that can respond. That is how we build GALE, with AI built into diagnostics, telemedicine, and logistics, supporting clinicians rather than replacing them.

The ministers we met are asking the right questions. Where does it run when the connection drops? Who is accountable for the decision? How is it paid for after year one? Who trains the people using it?

What comes next

Soon we will publish our view on a pragmatic way for governments to deploy AI in health: where to start, what to put in place first, and how to move from pilot to national scale without waiting for perfection.

The technology is here. Leadership is what makes it real.