The ambient AI story has moved on, and health systems are now building what comes next
A field report from “Enterprise-Grade AI as Infrastructure: Scaling the New System of Intelligence for Healthcare,” the Abridge summit at Becker’s 16th Annual Meeting.
For the last three years, the story of ambient AI in healthcare has been remarkably consistent. A clinician stops typing and starts listening. Pajama time declines. Notes are completed faster. Patient experience improves and clinicians make it home for dinner.
That progress matters. But it addresses only the surface of a deeper problem. Healthcare still requires roughly two hours of administrative work for every hour of care, contributing to burnout, delays in reimbursement, and nearly $1 trillion in annual waste.
The more important question, and the one that shaped the summit, is what happens when the clinical conversation itself becomes the source of truth. Every workflow in healthcare—documentation, coding, quality, revenue cycle—ultimately traces back to that conversation. Until now, it has never been captured, structured, and operationalized in real time.
The answers had less to do with documentation itself and more to do with what becomes possible when clinical conversations are transformed into a real-time system of intelligence.
One signal, many systems
The architectural shift Abridge is driving was put most plainly by Muhammad Siddiqui, the chief digital and information officer at Reid Health. The 245-bed rural system, located 65 miles from Indianapolis, serves 285,000 patients, the majority covered by Medicare and Medicaid, and has partnered with Abridge for more than two years.
Like many rural providers, Reid operates with tight margins, often in the single digits, leaving little room for inefficiency. It also faces the same capacity constraints as much of the industry, without the ability to simply hire its way out of the problem. Primary care appointments are booked 27 days out, and specialty visits stretch four to six months. In that environment, adopting AI is not about innovation for its own sake — it’s about expanding access and making the system work with the resources available.
The Abridge platform, Mr. Siddiqui told the audience, has become the “wiring” of the health system — a real-time intelligence layer flowing from the clinical conversation into every downstream workflow. This is the architectural shift: not software that produces documentation, but a system of intelligence that operationalizes the clinical conversation itself.
Abridge launched as an ambient documentation solution and has evolved into a real-time system of intelligence anchored on the clinical conversation — a single input that feeds documentation, coding, quality, and revenue cycle simultaneously.
The distinction matters. A documentation tool saves time. A system of intelligence changes how the work gets done. At Reid Health, that has meant a roughly 6% lift on the NRC patient-experience question about whether the clinician was paying attention, a 4% increase in work RVUs among clinicians using Abridge versus those who were not, improved HCC capture for Medicare Advantage, fewer CDI queries, and a drop in outpatient denials.
The change-management implications are their own kind of evidence. IT rollouts in healthcare are notoriously slow, contested, and mired in committees. Reid Health’s was not. “Within a week people are knocking on my door: ‘When can you implement this in my clinic?'” Mr. Siddiqui said.
What clinicians do with returned agency
Clinician leaders throughout the summit agreed: ambient AI is not just about what it lifts from clinicians, it’s about what it gives back.
Andrew Narcelles, MD, medical director of clinical informatics at OhioHealth, who oversaw the Abridge rollout across the 16-hospital system, described that return in cognitive terms. “I’ve been doing this for more than 20 years,” he said. “I interact with my patients all the time, but I’m constantly going back and forth between the EMR to document, to put a quick note so I don’t have to remember that later. That little bit of interaction distracts me from seeing the patient and interacting with the patient. That’s gone. That’s the amazing part because now I can just listen to the patient, talk to the patient, do all those things that I need to do,” said Dr. Narcelles.
What’s being restored is not just time; it’s attention. The return of agency is not a downstream effect of documentation relief. It is the point. Brandon Harris, MD, associate CMIO at UI Health, captured that sentiment in one of the most quoted lines of the morning: “We don’t want someone to tell us what to do. We know what we want to do. Unfortunately, the system usually gets in our way. And when we don’t know what we want to do, we want someone to help us in that moment.”
As Guru Sundar, VP of marketing at Abridge, said in his closing remarks: “As infrastructure evolves, it creates a world of agency.”That idea surfaced in every session — not as a slogan, but as the connective tissue between architecture and outcome. A platform built to produce notes will produce notes. A system of intelligence built on the clinical conversation does something fundamentally different: it restores agency. Because when intelligence is delivered in real time, at the point of care, clinicians are no longer navigating the system — the system is working with them. Time, attention, and decision-making return not by accident, but by design.
Clinical decision support, shaped by the conversation
If ambient documentation was the first use case, Abridge’s integration of clinical decision support into the conversation itself is one of the clearest expressions of this system-of-intelligence model. The work is built on a partnership with Wolters Kluwer’s UpToDate, announced in October 2025. The platform surfaces evidence-based guidance inside the conversation in real time, directly within the clinical conversation, without requiring clinicians to leave the encounter.
The design challenge is not trivial. Clinical decision support has a long and largely unsuccessful history in healthcare IT, defined by the alerts that go unread and interruptions that pull clinicians further from the patient rather than closer. Steven LoBue, MD, physician director for healthcare informatics at Northwestern Medicine, has spent years building and refining these systems. His conclusion is blunt: “I don’t think I’ve ever had anybody thank us for a new alert.”
Conversation-driven clinical decision support starts from a different premise. It aims to avoid the category’s original sin: surfacing guidance based on what the system happens to know, rather than what the clinician is actually trying to solve. Traditional decision support reacts to available data. Conversation-driven intelligence responds to clinician intent. “When it works, you have those moments where the question it surfaces is exactly the one you were already thinking about for that patient,” Dr. LoBue said. “That’s where you start to see real delight, the kind that can actually overcome burnout.”
Abridge is not replacing clinical judgment. It is sparing clinicians the hunt for evidence they already know how to apply, at the moment judgment is being exercised rather than after. The roadmap extends beyond external content to include system-specific pathways, protocols, and payer logic, bringing historically downstream decisions into the moment of care.
Dr. Harris put the long-view version more plainly. “I don’t think AI can replace us,” he said. “It can definitely help us, but I think we’re going to be physicians as long as we want to be.”
Trust and partnership as the preconditions for scale
Every panelist pointed to the same prerequisite: transparency. Dr. Narcelles wanted to see the transcript behind a note, the references behind a suggestion, and the model’s account of the question it thought he was asking — not as a compliance requirement but as the foundation for trusting any output at all. This is where features like Linked Evidence, the ability to trace any output back to the source conversation, become critical to closing the trust gap in AI.
The choice of source matters as much as the architecture around it. A reference clinicians already use is one they do not have to learn to trust. “We’re always looking for ways to cut down on clicks and help our clinicians spend more time with patients,” Veena Jones, MD, chief medical information officer of Sutter Health, said. “Abridge and UpToDate are two platforms our teams already use and trust.”
Dr. Harris, of UI Health, made the more structural argument: trust in AI at this velocity of development is not something any health system can establish alone. “We’re really good at delivering care, we’re okay at implementing technologies, and we try to get better every day,” he said. “We’re definitely not experts at building and innovating on AI tools. So we have to lean on our partners. And having good partnerships really allows that to be as seamless as possible.”
What is worth noticing about Abridge’s partner list, taken as a whole, is its heterogeneity. The company now supports more than 250 health systems, processing tens of millions of clinical conversations annually. Rural community hospitals like Reid sit alongside large academic medical centers, integrated delivery networks, federally qualified health centers, and specialty cancer hospitals. Those organizations differ in payer mix, margin, geography, and specialty concentration in almost every way a health system can differ. The structure of the clinical conversation — the foundation of care — does not.
Nursing, rev cycle, and what else the conversation can power
Reid Health has extended the platform from its physicians to its nursing staff, beginning with one floor and rolling out across seven more in the weeks following the summit. “Everybody wants to turn that on,” Mr. Siddiqui said.
The nursing work, built in partnership with Mayo Clinic and Epic, is a clear sign that the conversation-as-infrastructure architecture goes beyond physicians. Nurses document in discrete forms and flow sheets rather than narrative, and off-the-shelf ambient tools have not traveled well across that distinction.
The revenue cycle effect is structurally the most interesting piece. When the conversation becomes the source of truth, financial and operational workflows improve as a direct byproduct, not as separate processes. Problems that would have gone unbilled get billed. Denials fall. A patient presents with five problems; a clinician documents all five, not just the three they had time to capture manually. “That’s where the revenue cycle lift comes in,” Mr. Siddiqui of Reid Health said.
The ambient AI era made one thing clear: documentation does not have to come at the expense of clinician burnout. That question is settled. What comes next is more fundamental. Healthcare is beginning to build a real-time system of intelligence on top of the clinical conversation itself, one that connects care delivery, operations, and financial performance at the source. The question is no longer whether the clinical conversation can carry more weight. It is how much more, and which health systems will operationalize it first.
At the Becker's 11th Annual IT + Revenue Cycle Conference: The Future of AI & Digital Health, taking place September 14–17 in Chicago, healthcare executives and digital leaders from across the country will come together to explore how AI, interoperability, cybersecurity, and revenue cycle innovation are transforming care delivery, strengthening financial performance, and driving the next era of digital health. Apply for complimentary registration now.