Can AI Actually Solve the Healthcare Workforce Shortage?

I sat across from a hospital administrator in Riyadh last year who told me, without any prompting from me, that her institution was not actually short on doctors relative to patient volume, it was short on doctor-hours that were free to see patients. The nurses and physicians on her roster were fully employed and fully exhausted, but a large share of what they were doing in any given shift was not clinical work at all. It was chasing down prior authorizations, re-entering the same patient information into three different systems that did not talk to each other, and fielding routine questions that had nothing to do with the reason the patient actually needed a clinician. Health workforce researchers have been warning for years about a global shortfall commonly estimated in the tens of millions of workers, and that number is real and worth taking seriously, but it obscures a more immediate and more solvable problem, which is that the clinicians we already have are spending a large fraction of every shift on work that does not require a medical license.

This distinction matters because it changes what the right AI intervention looks like. If you believe the shortage is purely a headcount problem, you go looking for AI that pretends to be a clinician, and you end up disappointed or, worse, you end up with a system making judgment calls it has no business making. If you believe the shortage is substantially a time-allocation problem, you go looking for AI that clears the non-clinical work off a clinician's desk so the hours they do have go further. I have come to believe strongly in the second framing, not as an article of faith but because it is the framing that survives contact with an actual hospital's operations.

When we sat down with clinical teams during early product work, we asked a deliberately unglamorous question: walk me through your last ten patient encounters, minute by minute, and tell me where the time went. The pattern that came back, across primary care clinics and specialist practices alike, was strikingly consistent. Roughly a third of the time in and around a visit went to documentation, much of it duplicative. Another meaningful share went to coordinating between systems, chasing lab results that existed somewhere but were not visible where the clinician needed them, or re-explaining a patient's history because the prior record had not been carried forward cleanly. The actual differentiated clinical reasoning, the part that requires years of training and cannot be delegated to anyone or anything, was often a minority of the encounter's total time cost.

None of that is a criticism of any clinician or any single hospital. It is what happens when clinical systems grow by accretion over decades, layering new requirements, new payers, and new software on top of workflows that were never redesigned end to end. The opportunity is not to replace the third of the visit that is genuinely clinical, it is to compress the two-thirds that is not, and give that time back to patients rather than to paperwork.

The specific place I have seen AI earn its keep in staffing-constrained settings is not diagnosis, it is triage and routing, which are related to diagnosis but categorically different in what they demand and what they risk. A well-built intake tool can gather a patient's symptoms, history, and urgency signals before a clinician ever enters the room, structure that information consistently, and flag the cases that need to be seen sooner rather than later. That is not a diagnosis, and it should never be presented to a patient as one. It is a sorting function, the same kind of function a well-trained triage nurse performs, done at a scale and consistency that lets a smaller clinical team keep pace with a larger patient volume without anyone's safety margin shrinking. The clinician still makes every clinical judgment. What changes is that they make it with better-organized information and without having spent fifteen minutes gathering it themselves.

I want to be direct about the limit here, because I think overclaiming on this point is where the industry does real damage to trust. AI absorbing administrative burden and triage does not mean AI is closing the workforce gap on its own, and anyone who tells a ministry of health that it will is setting that ministry up for disappointment. What it means, more modestly and more defensibly, is that a health system with a fixed number of clinicians can extend the reach of each one meaningfully, particularly in the parts of the patient journey that are structured, repetitive, and low-ambiguity. That is a real and valuable outcome. It is just not the same outcome as solving the shortage.

At Hayya Med AI, this is the problem we built around rather than the diagnostic-replacement narrative that gets more headlines. Our tools are designed to sit ahead of and around the clinical encounter, not inside the clinical decision itself, handling intake, structuring patient history, and routing so that when a clinician's time is spent, it is spent on the parts of care that actually need a trained human being. We work in health systems across the GCC where the staffing math is unforgiving and where importing a Western-context workforce solution rarely fits the local reality of clinic density, patient volume, and language. The lesson from that work has been consistent: the systems that get real value are the ones that used AI to protect their clinicians' time, not the ones that tried to use it to avoid hiring clinicians at all. Those are different goals, and only one of them is honest about what the technology can currently do.