Across the Global South, hundreds of millions of people live without reliable access to basic healthcare. Too few doctors, too little data, and systems too fragile to scale. However, as Capria Ventures’ Francis Perelman outlines, a new wave of AI-native startups is building for precisely those conditions, deploying tools that would be redundant in wealthy markets but are decisive where the alternative is nothing.
In the developed economies, a chest X-ray is read within hours by a radiologist working nearby. In a rural clinic in a low-income country, the same scan can wait days for a qualified reader, if one can be reached at all. That distance is not a matter of a few years of catching up. It is structural, and it shapes daily medicine for a large share of the planet. It is also, increasingly, the gap a generation of founders has set out to close.
A Gap Measured in Billions
The World Economic Forum estimates that 4.5 billion people still lack access to essential health services, and projects a shortfall of 11 million health workers by 2030, a deficit that sits at the heart of the stalling global effort to reach universal health coverage by the same year. Sub-Saharan Africa carries roughly a quarter of the global disease burden while employing about three percent of the world’s health workforce, according to Brookings, which counts around one doctor for every 3,000 people in the region, a third of the ratio recommended by the World Health Organization. Specialist care is scarcer still. In some emerging markets there is one radiologist for every 100,000 people, against roughly one per 10,000 in high-income countries, and the shortage gets worse as the clinicians who qualify often emigrate to better-paid systems abroad.
This is not a market waiting to mature into the Western model. It is a different starting point, and it is the reason a technology built elsewhere may matter more here than anywhere.
When Software Learned to Reason
Artificial intelligence belongs to a small category of inventions that change the cost of an entire class of work, the way electricity and the internet once did. Over the past three years it has crossed a threshold. The narrow AI of a decade ago, built to flag a single feature in a single kind of image, has given way to general-purpose systems that read and write language fluently, recognize patterns in images and data at scale, draw inferences from incomplete information, and shoulder routine cognitive and administrative tasks that once consumed hours of skilled time.
What separates this moment from earlier waves of software is economics. These systems are becoming cheap, they improve quickly, and they run on commodity hardware or reach a user through an ordinary smartphone. That is why AI is spreading through finance, logistics, law and customer service faster than almost any technology before it. The more interesting question for healthcare is not how large the tool becomes, but who puts it to work, and where. The answer, more and more, is a new class of companies built around the technology from the start.
Expertise À La Carte
Much of the excitement around AI in the life sciences sits at the discovery end, where startups and large pharmaceutical companies alike are racing to design new drugs. A second frontier, quieter but no less consequential, lies in the delivery of care, where founders are turning the same technology toward patients who today receive little or none.
What makes the technology matter here is not any single application but the constraint it eases. Specialist expertise has always been tied to a person who is scarce, costly and clustered in a handful of cities, and out of reach for most patients beyond them. A capable model loosens that constraint. It lets a specialist’s input, or a usable approximation of it, reach places no specialist can be, at a cost per patient that falls as volume grows. The aim is not to do without clinicians but to extend the few who exist, with a trained professional reviewing and owning the decisions that matter.
It also reaches across the whole of care rather than one corner of it. The same underlying systems can support prevention and screening, inform a diagnosis, guide a course of treatment, follow a patient afterward, carry the administrative weight that sinks small facilities, and flag the early signs of an outbreak across a population. Each use tends to generate data the next can learn from, so a system improves as it runs rather than standing still. For founders, that breadth is the opportunity. What these markets have lacked is rarely a vision of better care, but the people, capital and information to deliver it, and those are precisely what the technology now supplies.
Built for the Shortfall
The value of a tool that extends expertise rises as that expertise grows scarce. In a hospital with ten radiologists on staff, an AI pre-read trims minutes from a workflow. In a region with one radiologist for every 100,000 people, that same pre-read can be the difference between a timely diagnosis and none at all. A capability that looks incremental in a well-resourced system can be decisive where the baseline is scarcity.
A 2024 report backed by the Novartis Foundation and Microsoft argued that lower-income countries could leapfrog wealthier ones by building AI-enabled health systems rather than slowly recreating the resource-intensive model of wealthier nations. Early evidence supports it. Rwanda, where a single doctor may be responsible for tens of thousands of people, now runs digital consultation services used by a large share of its adult population. Programs in Malaysia, Brazil and the Philippines use AI to forecast outbreaks of mosquito-borne disease.
The startups turning these capabilities into businesses are multiplying. In India, 5C Network runs an AI-assisted teleradiology service that pairs software with a network of radiologists who review and sign off on every report, returning results in minutes rather than days. In Nigeria, Helium Health operates electronic records for thousands of facilities and applies AI to surface population health patterns, while Argentina’s Lucai Bio uses AI-powered bioinformatics to accelerate drug discovery and biotechnology research. Neither is a household name, and in a sense that is the point. The most consequential work is happening quietly, founder by founder, in markets where these tools are not competing with an abundant supply of specialists but filling a gap where none ever existed.
A Bet Worth Making
The pattern these companies share is worth more than any single product. Some turn scarce expertise into a service reachable by anyone with a connection. Others build the records, billing and financing infrastructure that let a fragile system function and, in time, learn from its own data. None of this would stand out where specialists and infrastructure are already abundant. All of it matters where the alternative is insufficiency or nothing.
The caveats are real and worth stating directly. Models trained largely on Western populations can carry blind spots when applied elsewhere, which makes local validation essential and raises familiar ethical questions around bias, privacy and oversight. Connectivity is uneven, regulation is still developing, and reimbursement pathways are thin. There is a genuine risk of marketing outpacing evidence, a recurring pattern in digital health. None of this argues against the approach. It argues for capital patient enough to fund proper clinical validation, for more founders building specifically for these markets rather than adapting products designed for others, and for locally owned data.
What makes the moment unusual is that the technology and the need have arrived together. The capabilities now spreading through the global economy are, in much of the world, a convenience layered on systems that already work. Where the alternative is a scan no one reviews and a record kept on paper or not at all, the same tools are close to everything. The founders moving first are not waiting for the gap to close on its own. They are building where it is widest, and the case for backing more of them has rarely been stronger.
