THE QUIET POWER
How ambient AI is redefining healthcare workflows
One of the biggest lessons the team at Medical University of South Carolina (MUSC) learned about ambient AI (artificial intelligence) was that installing the tool wasn’t enough. When they started their ambient journey more than a year ago, the team had to spend time training the technology to understand medical terminology, local phrases, even accents.
Crystal Broj admits that the process takes commitment. But here’s the interesting part: Over the course of the one-year pilot, Broj, Chief Digital Transformation Officer, says the technology itself evolved so quickly that providers who started with MUSC noticed a huge difference by the end. “It got smarter, more accurate and much less disruptive. When we finally polled those clinicians, the overwhelming response was, ‘Don’t take this away from us.’”
The shift from skepticism to real attachment underscored a critical lesson in that adoption requires investment in both people and process. While the MUSC began with a small pilot, it eventually expanded to more than 100 providers, asking each to use the tool for a full year. Over time, a host of provider champions emerged who could explain the real-world benefits to their peers far more convincingly than the technology team.
To support them, MUSC introduced “white glove” training, pairing each user with an ambient coach to answer questions, take feedback and guide improvements. The strategy worked. By the end of the pilot, MUSC not only went system-wide, but also had a waiting list of more than 400 providers ready to join. “Ambient isn’t for everyone, and that’s okay,” Broj says. “The key is finding where it fits, supporting those who benefit most and respecting the fact that some workflows may not be a good match.”
The MUSC case study is a reinforcing example of the quiet power of ambient AI. The technology is transforming how healthcare organizations operate, offering not just efficiency but also the possibility of deeper human connection in the exam room.
For example, for Broj and her team, the transformation transcended something as simple as turning speech into text. “When people think of ambient tools, they picture transcription. But today these tools structure the note, suggest the right codes and even prepare billing. It’s the difference between having a tape recorder in the room and having a digital colleague who takes work off your plate.”
Integration with the electronic health record (EHR) also was critical to success. “If a tool doesn’t work smoothly in the EHR, clinicians won’t use it—period,” Broj says. “The best approach is API-driven connections that bring the AI right into the native workflow. No extra screens, no toggling between apps. When aligned, the experience feels natural and clinicians see big wins like faster charge capture, cleaner notes and fewer clicks.”
“When people think of ambient tools, they picture transcription. But today these tools structure the note, suggest the right codes and even prepare billing.”
— Crystal Broj, Chief Digital Transformation Officer, Medical University of South Carolina
Trust was another essential piece. MUSC created a dedicated AI governance committee to review new tools. MUSC didn’t just review security, but also appropriateness, the models being used and how the data was being handled. “Patients deserve transparency and the ability to opt out. Clinicians need confidence that the system isn’t watching them inappropriately. That governance builds trust.”
Overall, MUSC has tracked a 25%–40% reduction in “pajama time,” faster documentation closure, improved charge lag and fewer coding denials. Patients also reported feeling more connected, saying their doctors looked at them instead of their computers. “The beauty of ambient AI is that it works quietly in the background,” Broj says. “It doesn’t shout for attention—it just lifts the weight off clinicians so they can focus on care. That’s the promise: not just better notes, but better moments.”
Evolution and Embedded Adoption
At UW Health and the University of Wisconsin School of Medicine and Public Health, Joel Gordon, MD, views ambient AI as the long-awaited answer to a decades-old challenge—one that can help ease the relentless clerical load on clinicians. After years of trialing macros, smart tools and speech recognition systems that never scaled, a breakthrough has finally arrived.
“Reducing clerical burden has been the holy grail we’ve been seeking for the last 20 years,” says Dr. Gordon, Chief Medical Information Officer. “The plug-and-play ‘magic’ of AI allows broad adoption across our staff. Most specialties now experience career-enhancing impact and providers report revolutionary workflow improvements the very same day they’re trained.”
Dr. Gordon views the technology evolving in stages: first, “conversation to text,” where spoken encounters become structured notes; second, “conversation to orders,” which is already beginning in early form; and third, next-generation tools that include coding support, diagnostic suggestions, and decision aids designed to improve both efficiency and quality of care.
The rapid pace of development presents its own hurdles. “The R&D cycle is fast, and so we’re constantly shifting project maps and resources to align with alpha and beta releases,” Dr. Gordon says. “That’s challenging in healthcare when budgets are tight and margins are slim.”
“Reducing clerical burden has been the holy grail we’ve been seeking for the last 20 years. The plug-and-play ‘magic’ of AI allows broad adoption across our staff.”
— Joel Gordon, MD, Chief Medical Information Officer, UW Health and University of Wisconsin School of Medicine and Public Health
Interestingly, Dr. Gordon says skepticism has been surprisingly minimal. Instead, UW Health has focused on getting tools into the hands of providers who request them and tracking results closely with data dashboards. These dashboards also highlight peer mentors in each location, creating a network of champions for subsequent adopters.
Integration has been a key differentiator. UW Health deliberately chose an embedded product within its EHR vendor’s mobile application. “It makes us a much more untethered provider pool,” Dr. Gordon says. “Being able to interact with patient information walking down the hall, in the breakroom, or even in the garden behind the clinic adds another level of workplace joy. Ambient listening is becoming a driver application for how we engage digitally.”
Strong governance ensures responsible adoption. “The level of scrutiny we’re applying is beyond any other documentation solution,” Dr. Gordon says. “With great power comes great responsibility and these tools are very powerful.”
Early results support the investment. Dr. Gordon points to forthcoming UW Health research demonstrating significant improvements in efficiency, workload reduction and clinical relevance. For Dr. Gordon and his team, the quiet power of ambient AI lies not just in reducing clicks, but in restoring time and focus to the patient encounter.
Ambient AI is evolving from background tool to trusted partner, helping ease workloads and improve experiences across healthcare. The opportunity now is to scale it wisely, ensuring its quiet power supports both clinicians and patients in meaningful, lasting ways.
SIDEBAR
Ambient AI Playbook
UW Health recently published a protocol paper in The New England Journal of Medicine AI describing its framework for deploying ambient AI in clinical care. The goal was to reduce documentation burden while maintaining accuracy, compliance and patient trust.
Called the “Pragmatic Trial Operations (PTOps) Playbook,” the framework includes two phases. The implementation phase focused on governance, workflow integration and real-time monitoring. The trial phase was designed to measure provider well-being and efficiency outcomes.
Integration with the electronic health record used FHIR standards, while multidisciplinary workgroups applied the Systems Engineering Initiative for Patient Safety (SEIPS) model to ensure human factors and organizational alignment. Real-time dashboards tracked utilization and accuracy, and ICD-10 coding compliance was audited with a UW-developed large language model validated against professional coders. The stepped-wedge trial enrolled 66 providers across eight specialties, providing one of the first rigorous, evidence-based evaluations of ambient AI in practice.
