Ambient AI in Healthcare: Adoption & Challenges
- A recent study indicates that while ambient AI tools are gaining traction in healthcare, other clinical and operational applications face significant adoption challenges.The research, led by Eric poon,...
- The study, published in the Journal of the American Medical Informatics Association, found that 100% of respondents were developing, piloting, or deploying ambient notes, a generative AI tool...
- This near-global adoption of ambient notes contrasts sharply with other AI applications, including those for imaging, diagnosis, risk stratification, revenue cycle management, and research.Ambient notes was the only...
Ambient AI is rapidly transforming healthcare, yet adoption is uneven. A new study reveals that while ambient notes are widely embraced for clinical documentation, other applications face significant hurdles. explore how tool immaturity and inconsistent health equity evaluations impede the broader implementation of AI in healthcare settings across the U.S. Conducted by Duke University Health System, this research, published in the Journal of the American Medical Informatics Association, shows that while 100% of health systems use ambient notes, other areas lag. Discover the barriers health systems face, like financial concerns and regulatory uncertainty, and how evaluating performance, safety, and equity can definitely help. news Directory 3 highlights the need for practical, shared approaches to AI evaluation. Discover what’s next.
Health Systems Embrace Ambient AI, but Broader Adoption is uneven
Updated June 3, 2025
A recent study indicates that while ambient AI tools are gaining traction in healthcare, other clinical and operational applications face significant adoption challenges.The research, led by Eric poon, MD, MPH, chief health details officer at Duke University Health System, surveyed 43 U.S. health systems in the fall of 2024.
The study, published in the Journal of the American Medical Informatics Association, found that 100% of respondents were developing, piloting, or deploying ambient notes, a generative AI tool designed to draft clinical documentation from patient-clinician conversations.More than half reported a high degree of success with this technology.
This near-global adoption of ambient notes contrasts sharply with other AI applications, including those for imaging, diagnosis, risk stratification, revenue cycle management, and research.Ambient notes was the only request to achieve such extensive engagement, suggesting rapid acceleration in the documentation domain.
A key driver of this trend may be the emphasis on reducing clinician burnout. According to the study, 72% of organizations ranked relieving caregiver burnout and improving satisfaction among their top two goals for adopting AI. Workflow efficiency and patient safety were also high priorities. One respondent noted that alleviating documentation demands has become critical for retaining clinical staff.
Barriers to Clinical AI Integration
Beyond ambient notes, the study revealed inconsistent adoption and limited success of AI across other clinical and operational areas. While 90% of surveyed organizations had implemented imaging and radiology tools to some extent, only 19% considered the outcomes highly successful.
Clinical risk stratification tools, such as those for early sepsis detection, were adopted by about half of the respondents, with only 38% reporting high success. Similar patterns were observed in revenue cycle and diagnostic applications, where perceived effectiveness remained modest.
Tool immaturity appears to be a major obstacle. The study found that 77% of respondents identified it as a top barrier to AI adoption. Financial concerns (47%) and regulatory uncertainty (40%) were also significant factors. Low clinician uptake and lack of leadership support were less frequently cited,suggesting that technical and systemic challenges outweigh cultural resistance.
“AI is exciting, but the tools need to meet clinical expectations. right now, many of them simply don’t,” one executive said.
Evaluations and Equity
The study also highlighted that health systems are not consistently evaluating the performance, safety, or equity implications of AI tools. While most respondents routinely measured AI tool usage, only 17% always assessed health equity impacts, and 10% never evaluated equity at all.
This gap raises concerns about potential disparities. The study emphasized the need for shared evaluation frameworks, common deployment platforms, and better information exchange between vendors and providers.
“Practical,shared approaches to AI evaluation—before,during,and after deployment—will be critical,” the authors wrote.
Interestingly, health systems with higher net patient revenues (above $5 billion) did not significantly differ in their prioritization of goals or identification of barriers compared to smaller organizations, indicating that financial scale alone does not guarantee more advanced AI integration.
What’s next
The study urges organizations to collaborate on evaluation methods and share lessons learned to accelerate safe and effective AI use in healthcare.Without these shared guardrails, the authors caution, the industry risks deploying AI that does not serve patients or providers as well as it could.