You cannot recruit your way out of a 141,160-physician shortfall. You cannot hire your way past 17.6% RN turnover, either, not when the national vacancy rate sits at 8.6%, and every hospital in the country is fishing from the same shrinking pond. So if the workforce math doesn’t work, the only lever left is what we let our current people spend their time on.
You can’t hire your way out of this
The Health Resources and Services Administration projects a shortage of over 141,000 full-time equivalent physicians by 2038, with 30 of 35 specialties modeled expected to come up short. Nonmetro areas will feel it hardest: a projected 58% physician shortage, compared to 5% in metro markets. On the nursing side, the 2026 NSI National Health Care Retention & RN Staffing Report puts turnover at 17.6% nationally, with the average hospital losing $5.19 million a year to RN churn.
I used to think of these as staffing problems. In reality, they’re not. They’re people-of-healthcare problems, and the distinction matters. When a hospital can’t fill a role, that’s not just a line item on someone’s operations dashboard. It’s a patient who waits longer for discharge, a nurse covering two assignments instead of one, an aging population that needs more care exactly as the workforce meant to deliver it is shrinking. We are not going to produce enough new physicians and nurses to close this gap through the pipeline alone. The people already in the building are the whole strategy now.
Cost savings is the wrong question
Here’s where I think most health systems, and most of the vendors selling into them, get the AI conversation backwards. They pitch it, and buy it, as a cost-reduction tool. Fewer FTEs. Lower labor spend. A way to do more with less.
I’ve had this exact conversation with clinical leaders across several organizations we work with, on the payer side, the TPA side, and the hospital side. Every one of them has said some version of the same thing: the organizations getting this right aren’t using AI to shrink their teams. They’re using it to give their existing teams back their capacity. That’s a fundamentally different investment thesis, and it changes what you build, what you buy, and what you measure.
Augmentation, not reduction. AI in the background, handling the documentation, the triage, the administrative noise, so that the clinician in front of the patient can actually be a clinician. That’s not a cost play. It’s a capacity play, and capacity is the actual constraint standing between most hospitals and the growth or margin targets they’re chasing.
What are we actually measuring?
Most AI evaluations I see start and end with a dashboard that includes things like time saved on one workflow, error rate on one tool or a single metric, boxed in, disconnected from everything else happening around it. I’d argue that’s the wrong way to look at the impact of AI entirely.
A dashboard without a human story attached to it doesn’t tell you much. What I want to know is the cumulative effect: Is this saving nursing staff real time, not just theoretical time? Is it freeing the administrative team to focus on higher-value work? Is it letting a specialist spend more of their day on the cases that actually require their training, instead of the paperwork around them? Are outcomes improving? Are patients more satisfied, or less?
None of those questions live in a single tool’s reporting box. They live in the ecosystem the tool sits inside. If a hospital is only measuring the box, it’s missing the point of the investment. The organizations that will win the next five years of this staffing crunch are the ones who evaluate AI by its effect on the people and patients around it, not by whether one dashboard trended in the right direction last quarter.
The real cost of getting this wrong
The NSI report puts a number on this that’s worth sitting with: every percentage point of RN turnover costs or saves the average hospital roughly $295,000 a year. That’s real money, and it’s the number that gets a CFO’s attention. But the bigger cost is actually the one that doesn’t show up on a P&L.
If we treat this as purely a business-of-healthcare problem, we miss what’s actually at stake. Burnout drives turnover. Turnover drives vacancy. Vacancy drives longer waits, thinner coverage, and more strain on the people who stay. That strain compounds, especially for an aging population that needs more from the system every year, not less. This is a societal problem wearing a business suit. Hospitals that only solve for the balance sheet will find the balance sheet keeps getting worse, because the underlying people problem never went away.
Where this leaves us
The next time your organization evaluates an AI investment, I’d push you to ask a different question than the one most vendors are prepared to answer. Not “what does this cost or save.” Ask: does this give our people back the ability to practice at the top of their license? Does it let a nurse be a nurse, a physician be a physician, an administrator focus on the work only a human can do? That’s the ROI that actually moves the needle on a workforce gap we cannot hire our way out of. Everything else is a rounding error.
