AI Is Now Quoted in FTEs. Your Headcount Request Has Competition That Already Published Its Math.

This week: the sentence a health system put in print that changed the unit nursing AI is measured in, what a third of your vacancy gap looks like when it arrives as software instead of a hire, and the gap between what your CFO has already read and what your nurses have been told.

By Christina Archer, Founder · Week of October 5, 2026

For three years the nursing AI conversation has been held in a currency your staffing plan does not use. Hours returned. Minutes saved. Satisfaction points. Nothing you could put in a headcount request, and nothing anyone could put in one against you.

That changed this year, quietly, in the measurement sections of reports written by health systems about their own programs. The number is now denominated in full-time equivalents, and it is in print.

▪One health system has already converted its AI program into positions, in writing

The MUSC Telehealth Center of Excellence published a virtual nursing measurement framework this year. Buried in the results is the sentence that matters to anyone who owns a requisition: 23,516 virtual nursing visits, at a mean visit length of 36.8 minutes, delivered care “equivalent to approximately 10.2 full-time nursing positions.”

Read that again as a hiring sentence. A health system measured its own AI-enabled program, converted the output into FTEs, and printed the number.

The program went from five units in late 2023 to 57 units across four regions by April 2026. Turnover in the early-adopter units fell from 47% in FY24 to 30% in FY25, and on one unit from 150% to 40%. MUSC puts the turnover savings in those units near $2.4 million.

None of that is a vendor projection. It is a provider reporting on itself, in a document written to help other systems measure the same thing.

▪The documentation number is now peer-reviewed, and it arrives per shift

Here is the second number, and it is the one your finance team will find first because it is the easiest to multiply.

A study published in Nurse Leader in August measured ambient documentation across five medical-surgical units at four Bon Secours Mercy Health hospitals, using 29,467 encounters pulled from Epic. At 60% ambient capture or better it found 20.6 minutes of recoverable time per twelve-hour shift. Apply the leading unit’s performance system-wide and the figure is 36.3 minutes. At early adoption, around 30% capture, it was 13.8 minutes.

The context the same study gives is why those minutes exist at all: nurses nationally spend between 23% and 40% of a twelve-hour shift documenting care — somewhere between 2.76 and 4.8 hours of every shift you staff.

Three capture rates, three numbers, one peer-reviewed inpatient nursing source. A year ago that sentence could not be written.

▪Convert the minutes into the only unit your plan is written in

Minutes do not compete with a requisition. FTEs do, so do the conversion before someone else does it for you.

Twenty-point-six minutes of a 720-minute shift is 2.86% of it. Across a hundred bedside RNs, on paid hours, that is 2.86 FTE-equivalents of time returned. At the leading-unit rate it is 5.04. At early adoption it is 1.92.

Now set that beside the hole you are hiring into. NSI puts the national RN vacancy rate at 8.6% — 8.6 open positions per hundred. So the published ambient documentation range covers between 22% and 59% of your vacancy gap, with a third of it as the middle case, and not one hire in it.

The comparator on the other side is your own funnel. An experienced med-surg RN req takes 83 days to fill, which is the specialty the Bon Secours study ran on. That is the choice as a CFO will frame it, and the framing is not unfair.

One honest note on the arithmetic, because it moves the answer. I converted on 2,080 paid hours per FTE, which is the conservative end. MUSC’s own 10.2 figure implies roughly 1,414 worked hours per FTE — a productive-hours convention. Run the same 20.6 minutes on that basis and the hundred-RN number is 4.2 rather than 2.86. Whichever convention your finance team uses for nursing FTEs is the one your conversion has to use.

▪Separate the programs that returned hours from the two that moved a labor line

This distinction is the single most useful thing in this edition, and almost every write-up blurs it.

Most published results are hours returned to bedside nurses. Advocate Health reported more than 43,000 hours across 25 hospitals in 2024, rising past 72,000 later in the year. Piedmont Healthcare reported 84,460 hours saved on discharge. WellSpan Health reported 4,300 hours across eight hospitals, about two hours per shift. Ochsner reported roughly 7,500 bedside hours; MD Anderson, 2,247; Henry Ford, more than a hundred hours a month.

Hours returned are capacity handed back to patient care. They are not positions you did not fill, and a hiring leader who lets the two be used interchangeably has lost an argument that was winnable.

Two results are different in kind. Advocate attributes $6.3 million in cost avoidance specifically to reduced nursing turnover. The Guthrie Clinic, in results reported in late 2024 on a program running since 2023, took its travel nurses from 172 to 98 full-time equivalents and put the labor saving at $7 million, with turnover falling from 25% to 13%. CoxHealth reported turnover falling from 25% to 11%.

Those are contract labor lines and retention lines, which is to say they are your lines. The through-route is not that software replaced nurses. It is that nurses stayed, and a hospital that keeps its nurses stops renting them.

▪Your nurses are being handed tools they do not trust, and that lands in your offer conversation

Set the deployment numbers beside what nurses say when an independent surveyor asks. Elsevier’s Clinician of the Future 2026: Nurses Edition surveyed 692 nurses between December and February.

Only 41% of nurses report using AI at work, against 57% of doctors. Only 42% say the tools are trustworthy today. Only 46% say their institution is good at providing AI training. And 41% say their views are rarely or never adequately represented in their organization’s decision-making — against 19% of doctors.

That last figure is the recruiting variable, and it is the one nobody is pricing.

The nurses are not refusing the technology. 80% say AI will not replace clinicians but will become a critical assistant, and 78% expect AI skills to be essential to clinician training. They are telling you they were not asked. Piedmont, on the other side of that, reports 85% of its nurses saying they love the program and 78% saying it makes them more likely to stay.

My read: AI tooling has become a benefit line, and it is currently being advertised by nobody. Candidates are three months from asking what a unit runs and who was consulted before it was turned on, and the hospitals that can answer that in an interview will convert offers against hospitals that cannot, at the same wage.

Our Workforce Intelligence dashboards carry vacancy, turnover and time to fill by specialty for the market you are hiring in, which is the other half of this conversion.

▪Five moves before your next headcount cycle opens

1. Get your own ambient capture rate and minutes figure from informatics this month. Somebody in your organization already has it, and right now it is a clinical informatics metric rather than a workforce one. The published range runs from 13.8 minutes at 30% capture to 36.3 at the leading-unit rate; where you sit inside that range decides what the conversion is worth.
2. Do the FTE conversion yourself, on your own hours-per-FTE convention. Minutes divided by 720, times your bedside RN count, divided by your own paid or worked hours basis. Bring it to the budget meeting rather than meeting it there. The conversion is going to be done either way, and the person who does it first sets the terms.
3. Put “hours returned” and “positions avoided” in separate columns, in writing. Nearly every published figure is the first. Two of them — Advocate’s $6.3M and Guthrie’s 172-to-98 — are the second, and both got there through retention rather than through substitution. That distinction is your whole position.
4. Recruiters: ask what the unit runs and who was consulted before it went live, and then use the answer. Seventy-eight percent of Piedmont’s nurses said the program makes them more likely to stay. Forty-one percent of nurses nationally say they are never asked. A unit that can describe its tooling and its nurse involvement has an offer advantage, and nobody is putting it in a posting.
5. Staffing and recruiting firms: the systems publishing FTE-equivalents are the ones repricing their contingent lines. Guthrie went from 172 travel FTEs to 98. A client with a measured virtual nursing program is a different account from one without, and the ones that publish their numbers are telling you which they are.

▪The unit of account changed, and nobody told talent acquisition

Every hiring leader I talk to is measured on time to fill, cost per hire and offer acceptance. None of those three move when a hospital recovers twenty minutes a shift, which is exactly why the conversion is going to be presented by someone whose metrics do move.

The published evidence is better than the hype and narrower than the fear. It says hours come back, it says turnover falls where nurses were part of the rollout, and in two named cases it says a contract labor line shrank. It does not say anywhere that a req was cancelled and the care still happened.

Do the FTE math on your own numbers this month, and walk into the headcount conversation holding it.

Hiring right now? Posting a nurse-exclusive opening on The RN Network job board is free to start — post a job →