There’s something almost charming about watching a Big Tech executive sketch out the future of medicine whilst carefully not mentioning that half the world’s clinics still run on paper, intermittent electricity, and prayer. Dr Angela Shippy, AWS’s Senior Physician Executive and Clinical Innovation Lead for Healthcare, spent a considerable portion of AWS re:Invent making the case for AI-powered healthcare transformation. And to be fair, some of it sounds genuinely useful.
Shippy’s background gives her credibility that most tech evangelists lack. She’s an internal medicine physician who started her career with paper charts and books in pockets, survived the painful transition to electronic health records, and now finds herself bridge-building between Silicon Valley’s promises and the messy reality of actual patient care. When she talks about clinicians drowning in documentation, she’s describing a problem she’s lived, not one she’s read about in a McKinsey report.
The pitch itself is familiar: AI tools can handle the administrative slog that’s burning out healthcare workers, freeing them to actually practice medicine instead of becoming glorified data entry clerks. AWS HealthScribe, built on the company’s generative AI stack, promises ambient listening that automatically generates clinical notes whilst doctors maintain eye contact with patients. Jupiter Medical Center’s deployment of Epic on AWS, combined with an AI-powered call centre, eliminated an 1,800-patient backlog for procedures and radiology whilst achieving 95% equipment utilisation. These aren’t theoretical gains.
But then there’s the other conversation, the one about deploying these same tools in underserved and rural communities where “cloud-first” might be a bit optimistic when “electricity-first” hasn’t been fully sorted. Shippy acknowledges the barriers exist, then pivots to AWS’s $60 million health equity initiative and partnerships like Huron AI, which helps 15 oncologists in Rwanda manage a population of 13.5 million through a text-based app. It’s a clever example, partly because it sidesteps the infrastructure question by using SMS instead of demanding high-bandwidth connections.
The tension becomes more obvious when you consider what healthcare AI deployment actually requires. You need reliable connectivity, structured data systems, trained personnel, and the kind of digital infrastructure that many regions simply don’t have. Shippy’s response is that AWS starts with level 100 foundational courses and builds from there, democratising data and ensuring communities have access. Which is true, but also somewhat evasive. Teaching someone cloud fundamentals doesn’t solve for the clinic that loses power three times a week or the rural hospital where “high-speed internet” means a 3G connection on a good day.
AWS’s case study library includes Jacaranda Health, which uses the company’s infrastructure to improve maternal and infant care outcomes in Nigeria, Kenya, and Ghana. The organisation combines SMS messaging, voice calls, and a mobile app to reach new mothers in communities where traditional healthcare infrastructure remains patchy. It’s the kind of deployment that works precisely because it’s designed around existing constraints rather than demanding new infrastructure first. The model acknowledges reality: you meet people where they are, with the technology they already have access to, rather than waiting for perfect conditions that may never arrive.
The data sovereignty conversation reveals similar friction. Shippy insists AWS and sovereignty go together, that customers control where their data lives and who accesses it, that everything’s contractually bound and secured. But when pressed about the historical mistrust between Global South institutions and large tech companies, the answer stays safely in the realm of technical capabilities rather than addressing the underlying power dynamics. Yes, AWS has global infrastructure and robust security. That doesn’t necessarily resolve concerns about who ultimately benefits when healthcare data flows through American cloud providers, or what happens when commercial interests conflict with public health needs.
There’s also the question of who actually benefits from these AI-powered workflows. Shippy’s vision is admirably upstream-focused: use predictive modelling to identify at-risk populations, shift from managing chronic disease to preventing it entirely, get people the interventions they need before they’re patients. It’s the kind of public health thinking that sounds transformative, assuming you can aggregate enough clean, structured data to make the predictions meaningful. Which brings us back to that statistic Shippy mentions: 97% of healthcare data goes unused because it’s unstructured. Solving that problem in well-resourced health systems is hard enough. Solving it in settings where basic record-keeping remains inconsistent is something else entirely.
The Capitec example mentioned by Shippy offers an interesting parallel. AWS’s partnership with the South African bank isn’t about incremental improvement but fundamental transformation, turning a financial institution into something more like a technology company. CTO Andrew Baker’s framing resonates: trust used to mean physical branches and transparent fees, now it means always-on availability and data security. Healthcare faces a similar shift, where trust increasingly depends on digital infrastructure working reliably. But unlike banking, where you can choose your provider, healthcare access is often determined by geography, economics, and government policy. The stakes are different.
What Shippy describes as her ideal deployment reveals both the promise and the limitation of the approach. She’d give everyone a device at their fingertips, something that empowers them to know more about their own health, that sends preventive reminders about blood pressure checks and family history screenings. It’s a vision of healthcare that’s proactive, personalised, and data-driven. It’s also a vision that assumes universal device access, health literacy, and the kind of digital infrastructure that makes always-on health monitoring feasible. For some populations, that’s already reality. For others, it’s still fairly distant.
None of this means AWS’s healthcare AI work is misguided or that tools like HealthScribe won’t genuinely improve clinical workflows. They will, particularly in settings where the foundational infrastructure already exists. But the gap between what’s possible in a well-equipped hospital system and what’s deployable in resource-constrained environments remains substantial. Acknowledging that gap honestly, rather than smoothing over it with references to equity initiatives and skills training, would make the pitch more credible.
My article that examined how African nations are approaching AI regulation noted similar tensions between technological possibility and practical implementation. AWS isn’t unique in facing this challenge, but as one of the largest cloud providers pushing into healthcare, how it navigates the contradiction between “AI-first” ambition and “basics-first” reality will shape what digital health transformation actually looks like in practice.
Shippy’s closing thought about deploying something like Project Kuiper (now known as Amazon Leo) everywhere, giving universal access to health information, is genuinely aspirational. Whether it’s achievable without first solving for the infrastructure gaps that currently determine who gets access to quality healthcare remains the more complicated question. And perhaps the more honest one.