Seven in 10 U.S. hospitals now use predictive AI integrated with their electronic health record (EHR). That's according to federal data from the Assistant Secretary for Technology Policy (ASTP/ONC), which found adoption rose from 66% in 2023 to 71% in 2024. A separate 2026 survey of 120 health system executives by Eliciting Insights found 75% of systems now use at least one AI application, up from 59% a year earlier.
So the state of AI in hospitals heading into 2027 isn't really about adoption anymore. Most hospitals have started. The harder questions are about who's getting value, who's being left behind, and what it takes to turn a promising pilot into something that changes how a hospital runs every day.
Here are five trends that define where hospital AI stands right now, and what they mean for the leaders responsible for the most expensive, most constrained resource in the building: the operating room.
1. Adoption is no longer the question. Scale is.
Starting an AI project is easy. Finishing one is not. A 2024 RAND Corporation report found that more than 80% of AI projects fail, roughly twice the failure rate of IT projects that don't involve AI. A 2025 MIT study found that only 5% of custom enterprise AI tools make it into production.
Hospitals aren't immune. Many have AI tools that are technically live but haven't changed how decisions get made. The reasons are familiar: recommendations that live outside the daily workflow, unclear ownership, models that aren't recalibrated as practice patterns change, and data that isn't as clean as anyone assumed.
The organizations pulling ahead aren't the ones with the most AI pilots. They're the ones that pick a measurable problem, baseline it, and keep the tool in use long enough to prove it works. In 2027, "we have AI" will mean very little. "Here's what it changed, in dollars and minutes" will mean everything.
2. Hospital AI is splitting into two tiers
The most important finding in the federal data isn't the overall adoption rate. It's who is and isn't adopting. In 2024, according to ASTP/ONC:
- 96% of large hospitals used predictive AI, compared with 59% of small hospitals.
- 81% of urban hospitals used it, compared with 56% of rural hospitals.
- 86% of hospitals in multi-hospital systems used it, compared with 37% of independent hospitals.
- 90% of hospitals on the market-leading EHR used it, compared with 50% of hospitals on other EHRs.
That last gap matters most. Most hospital AI today comes from the EHR vendor: 80% of hospitals using predictive AI reported using models from their EHR developer. If your hospital isn't on the market-leading platform, or doesn't have a large system's IT budget, the default path to AI is much narrower.
This is a quiet equity problem in healthcare operations. Community hospitals, rural facilities, independent hospitals, and ambulatory surgery centers (ASCs) face the same OR pressures as large academic systems, often with thinner margins and smaller teams. They arguably need operational AI more, and they're getting it less.
3. Operations is where AI is proving its value
The fastest-growing uses of hospital AI aren't at the bedside. They're in operations. Among hospitals using predictive AI, ASTP/ONC found that use for billing automation jumped from 36% to 61% in a single year, and use for scheduling rose from 51% to 67%.
Clinicians are pointing in the same direction. In the American Medical Association's survey of nearly 1,200 physicians, 57% said reducing administrative burden through automation was the biggest opportunity for AI, more than any clinical use case.
That makes sense. Operational problems like scheduling, capacity, communication, and supply use are high-volume, measurable, and repetitive. They don't require an algorithm to make clinical judgment calls. And they have a clear financial signature, which matters a great deal right now.
For perioperative leaders, the implication is straightforward. The best near-term AI opportunities in surgery aren't in the sterile field. They're in everything around it: how cases are scheduled, how accurately their duration is predicted, how block time is released, how teams communicate on the day of surgery, and how patients and families are kept informed.
4. Budgets now demand proof, not promise
Hospital finances are improving unevenly, and 2027 brings new pressure. Fitch Ratings reported that the median operating margin for nonprofit hospitals it rates was 1.5% in 2025, below every pre-pandemic year in its 20-year dataset. Fitch expects Medicaid changes under the One Big Beautiful Bill Act to begin materially affecting providers in 2027.
Costs are rising faster than revenue. The American Hospital Association reported that total hospital expenses rose 7.5% in 2025, while hospital prices rose 3.3%.
Technology budgets are feeling it. In Sage Growth Partners' Health IT Purchasing Forecast 2026–2027, 41% of hospital and health system executives said they expect capital budgets to shrink, and 77% said anticipated return on investment is now the top factor in technology purchasing decisions, up from 50% two years earlier.
The bar for AI has changed. A tool that produces insight but not measurable financial results will struggle to survive the next budget cycle. Time to value, payback period, and evidence from a hospital's own data now matter more than feature lists.
5. Trust depends on data quality and governance
Hospitals are taking AI oversight more seriously. In 2024, 82% of hospitals using predictive AI evaluated their models for accuracy, 74% evaluated them for bias, and 79% conducted post-implementation monitoring, according to ASTP/ONC.
Governance can only go as far as the data underneath it. A 2026 study from the University of Wisconsin School of Medicine and Public Health, published in the Journal of Medical Systems, found a 17% error rate in manually documented OR timestamps, with error rates climbing above 25% when entries were delayed. Those timestamps are exactly what most OR prediction models are trained on.
The lesson for 2027: trustworthy AI starts with trustworthy data. Hospitals should ask any AI vendor how its models handle messy source data, how often they're recalibrated, and how accuracy is reported back to the people using them.
Why the OR is the proving ground for hospital AI
If operational AI is where hospitals will prove value, the OR is where the stakes are highest and the results easiest to measure.
OR time is expensive and getting more so. Every unused block, late start, and same-day cancellation has a dollar figure attached.
The OR is also a constrained resource. Hospitals can't easily add rooms, surgeons, or nurses. That means the gains have to come from using existing capacity better: more accurate case duration estimates, earlier block release, faster turnover, better coordination across pre-op, the OR, and post-anesthesia care. These are precisely the kinds of problems operational AI handles well.
And the OR touches nearly every part of hospital finance. Gains in perioperative efficiency show up in revenue, labor costs, supply spend, and patient experience, often within months rather than years.
What perioperative leaders should look for in OR AI in 2027
Given these trends, here's what separates OR AI that delivers from OR AI that stalls:
- It works with the EHR you already have. Not just the market leader. Hospitals on any EHR should be able to add operational AI without a replacement project.
- It fits your setting. Community hospitals, rural facilities, and ASCs have different workflows and resources than large academic systems. The platform should be built for them, not adapted as an afterthought.
- It specializes in perioperative care. The OR has its own logic: block time, case duration variation by surgeon, turnover, flip rooms, staffing. Depth beats breadth.
- It turns predictions into action on the day of surgery. A prediction no one sees in time doesn't help. Look for recommendations that reach schedulers, charge nurses, surgeons, and patients' families where they already work, including mobile.
- It connects operations to finance. The platform should show what improvements are worth, using contribution margin, case cost, and supply data, not just utilization percentages.
- It's honest about what AI can and can't do. Be wary of any vendor that promises AI will run your OR.
- It delivers value quickly. Look for modular deployment that proves one capability before expanding, rather than a multi-year transformation.
Where Leap Rail stands
We built Leap Rail around these principles because we've seen what happens when OR technology ignores them.
Leap Rail is a perioperative intelligence platform, and perioperative care is all we do. It layers on top of the EHR a hospital or ASC already uses, including Epic, Oracle Health, MEDITECH, Altera, and others, so community and independent facilities don't have to wait for their EHR vendor's roadmap to get operational AI. It deploys one module at a time, at your organization's pace.
Our approach is grounded in published research and real OR practice. Our founder and CEO, Shayan Zadeh, authored a chapter in the third edition of Operating Room Leadership and Perioperative Practice Management, published by Cambridge University Press, and we've written extensively on the operational practices that make ORs run better, including the SANDPO protocol, the Bullpen staffing model, and flip room strategy.
Most importantly, we believe the purpose of OR AI is to help the people running the OR make better decisions faster, not to replace them. That means accurate predictions, clear accountability, honest reporting, and results you can measure in your own data.
Want to see what operational AI could look like in your OR? Request a demo.