Arcadia, a healthcare data and analytics company, surveyed 281 provider, payer, and healthcare services leaders at HIMSS26 in March, according to Healthcare Finance News. Just over half, 52%, said AI can transform healthcare when applied well. Only 14% said AI insights are fully embedded in the decisions their organizations make. Another 53% said those insights reach daily workflows in part.

Six percent called AI overhyped, but overall confidence seems settled. The open question is where the output lands, and whether it makes it into real decisions.

Ask a nurse manager building next month’s schedule what changed after the AI pilot, and you often get the same answer: a dashboard exists, and the schedule still gets built the way it always was.

The obstacle leaders named first is a workflow problem

Respondents ranked embedding AI into routine decision-making as the biggest barrier to responsible adoption, at 31%. Educating executives and staff followed at 27%, data foundations at 22%, and measuring impact at 20%.

The largest things holding AI back in hospitals aren’t about the AI model being smart, but rather getting AI’s output into the real work where decisions happen.

AI might generate a dashboard or report, but the people making real calls (like scheduling nurses, approving overtime, or opening a job requisition) aren’t using it at the exact moment they decide.

Hospitals have already purchased/implemented AI tools, but the next challenge is making AI show up in the workflow so it actually changes what people do, especially for decisions with deadlines.

Leaders are already pointing the technology at labor

When asked which outcome matters most from AI investment, a third of respondents chose measurable cost savings. Reducing workforce turnover came second at 27%, and improving financial forecasting third at 21%.

If you look at the full picture of what healthcare leaders say they want AI to do (save money, reduce staff quitting, improve forecasting), they’re all really about labor, the people side of running a hospital.

The top three priorities point to one main idea: leaders expect AI to prove its value first by improving workforce decisions (scheduling, hiring, coverage planning) because they happen often, cost a lot, and you can see results quickly.

Forecasting is the honest test

A forecast only matters if it changes what you do next. If your census model says “we’ll need 6% more surgeries in October,” but the hospital doesn’t change staffing, schedules, or plans, then the forecast didn’t create value, it was just information sitting on a dashboard.⁠

We shared about this in our recent research report, where we feature insights from healthcare leaders on what a forecast is actually worth. The short version: a forecast pays off when three conditions hold. You can see the gap early enough to act. You have qualified clinicians who can absorb it. And you can measure what the fill costs you against what the gap would have cost.

Most health systems have the first condition covered. The second is where plans stall: a forecast only helps if you have credentialed clinicians ready to deploy.

If you’ve built a float pool or per diem bench during stable months, you can turn the forecast into filled shifts. If you haven’t, the forecast just turns into more overtime requests and a longer approval queue.

Why the gap persists

Timing is most of the difference. A coverage gap surfaced 10 days before the shift week is a planning problem. The same gap at 48 hours is an overtime bill. Nothing about the model changed between those two moments; what changed is who saw the number and whether they could act on it without a meeting.

The measurement question is simpler than it looks, too. Fill rate against forecasted need, and cost per covered hour against an overtime baseline, are enough to tell a CFO whether the forecast paid for itself. Systems that can’t produce those two numbers usually don’t have a model problem. They have a decision that never moved.

The bottom line

The 14% figure suggests most healthcare leaders aren’t actually using AI when they make important day-to-day decisions. Instead, AI usage in healthcare appears to be mostly in the testing or dashboard-viewing phase, not the “it changes what we do” phase.

Workforce planning is one of the clearest places to close that gap. These decisions repeat weekly, and with an accurate forecast, you can optimize your staffing mix before gaps turn into last-minute scrambles.