Table of Contents
Donna Roach had 100 licenses and no budget when she walked into her board retreat two years ago. She brought a demo of ambient AI technology anyway, and the board backed her instantly. That early bet on documentation technology has since evolved into a much bigger question about what it takes to be a lasting AI partner health systems can rely on as vendor relationships mature well beyond their original scope.
How Adoption Stopped Being the Question for Lasting AI Partner Health Systems
These were the arguments health systems were having about ambient AI for clinical documentation two years ago: does it work, and will clinicians actually use it? Across six systems examined in this reporting, UNC Health, BJC Health and WashU Medicine, Hartford HealthCare, University of Utah Health, NYU Langone Health, and WVU Medicine, that two-part question is now a case closed. Adoption isn’t the constraint anymore; neither is enthusiasm.
Two Shifts That Changed the Conversation
Two things have changed. The first is what Abridge is able to do, expanding from documentation into the full arc of the visit, including Care Signals, co-designed at Kaiser Permanente, which surfaces a patient’s key conditions before the appointment and closes them out in coded documentation afterward. The second is how health systems are grading the technology, with scrutiny increasingly focused on economics as some AI vendors move from predictable per-seat licensing toward utilization-based pricing tied to compute consumption, a variable cost no one has learned to forecast yet.
Adoption Curves Behind This Lasting AI Partner Health Systems Standard
UNC Health began a small pilot in August 2024 and expanded twentyfold between February and April 2025, renewing in February 2026 and now using 103% of its contracted seats, with more than 2.5 million notes generated as of June. “The level of adoption has been incredible, without having to really push it,” said David McSwain, MD, UNC’s system CMIO. “It is, I think, without question, the most impactful technology that I’ve ever deployed in my career.”
Organic Growth Across the Other Five Systems
WVU Medicine went live in May 2025 with about 50 clinicians and grew through “friends telling friends” to enterprise-wide status by November, recording more than a million notes. BJC’s pilot, begun in July 2024, has reached more than 2,000 clinicians. Hartford HealthCare signed an unlimited enterprise agreement in July, reaching more than 1,600 licensed clinicians. NYU Langone expanded from ambulatory practices to the emergency department to inpatient settings, now covering more than 1,400 regular users.
Measuring Financial Returns Within This Lasting AI Partner Health Systems Framework
BJC Health and WashU Medicine put just under 500 primary care providers on ambient AI and pointedly declined to measure whether it made money at first, betting that fixing clinician and patient experience would produce financial returns on its own. Note-writing time fell about 7% almost immediately, then roughly doubled to 15% by month five. Revenue came last: a 3% weekly lift in work RVUs among physicians, or close to $2 million a year across the group’s 200-plus doctors.
WVU’s Parallel Outcomes Data
At WVU, adopting the platform cut in-room human scribe reliance by over 30% through natural student attrition rather than layoffs. In clinician survey data supplied by Abridge, WVU users reported 61% less cognitive load, 77% more satisfaction at work, 78% more undivided attention to patients and a 43% gain in accommodating urgent cases.
The Pricing Uncertainty Testing Lasting AI Partner Health Systems Relationships
“Everyone is very anxious about trying to understand and plan for a shift towards utilization-based pricing models for AI,” said Philip Payne, PhD, BJC’s chief health AI officer. “I don’t believe anybody has solved the problem of accurate and reproducible forecasting of token utilization.” Systems are hedging differently: Utah put several million dollars into an on-premises Nvidia environment to keep token economics inside its own walls, while BJC chose a utilization-based budget model with micro-billing capability to track consumption at a fine grain.
Why This Timing Matters for Hospital Finances
This pricing shift is arriving as hospital operating margins run near 1%, payer mixes shift and the bulk of Medicaid reductions are set to land across late 2026 and 2027, adding financial pressure to an already complex technology procurement decision.
How Executives Test Whether a Vendor Qualifies as a Lasting AI Partner Health Systems Trust
Dr. Payne’s test is whether a vendor treats AI as a product with a life cycle spanning years rather than a static deployment. Barry Stein, MD, of Hartford HealthCare, tests for whether founders truly understand the clinical problem, technical depth beyond scribing, and robust AI testing and monitoring capabilities. For Roach, the proof is narrower: whether feedback changes something quickly, since she has “a very short and very small window to build up credibility with my clinicians.”
A Willingness to Be Told No
Adam Szerencsy, DO, of NYU Langone sets the bar at a vendor’s openness to pushback rather than defaulting to “we know better than you.” At UNC, McSwain developed his own test, watching whether a vendor’s leadership addresses the hardest questions about clinical workforce impact unprompted, something he found consistently true of Abridge.
Expanding Into Nursing and Clinical Decision Support Within This Lasting AI Partner Framework
Nursing documentation, which lives in structured flowsheets rather than narrative notes, represents a harder engineering problem and a far larger seat count if solved. WVU kicked off a nursing pilot in June on a single unit, with an October go-live planned. “Speaking and focusing on your patient and talking opens up your mind to think differently,” said Rebecca Mitchell-Perry, WVU’s inaugural CNIO.
Clinical Decision Support Moving From Population to Individual
Abridge’s context-aware clinical decision support draws on a content library anchored by The New England Journal of Medicine, JAMA, the CDC, and other sources. “Now we’re taking decision support from a cohort of patients at a population level and tailoring it to this specific patient in front of me,” Szerencsy said, noting the tool now also queues up orders and provides billing recommendations at NYU Langone.
What This Means for Health Systems Evaluating Their Own AI Vendors
Given the consistency across all six systems in describing similar evaluation criteria, technical depth, responsiveness to feedback, and genuine understanding of clinical problems, other health systems navigating their own ambient AI vendor relationships may find these standards a useful benchmark as they consider expanding beyond initial documentation pilots. Given the unresolved challenge of forecasting utilization-based AI pricing that Dr. Payne described, health systems should expect continued financial planning uncertainty as vendors shift away from predictable per-seat licensing models industrywide.
What to Watch Going Forward
As Abridge and similar vendors continue expanding from documentation into nursing, revenue cycle, payer adjudication, and clinical trial recruitment, industry observers will likely watch whether the trust built through initial ambient AI deployments transfers smoothly to these higher-stakes applications, or whether each new capability requires vendors to re-earn credibility as Roach described. Given the shared theme across all six systems that being a lasting AI partner health systems trust requires treating AI as an evolving product rather than a one-time deployment, this framework may increasingly shape how health systems structure vendor contracts and renewal decisions as ambient AI technology continues its rapid expansion across clinical and administrative workflows.
For more healthcare industry updates, insights and news, visit DistilINFO. Click here to subscribe to stay informed.
