Healthcare organizations are increasingly adopting artificial intelligence to improve efficiency and reduce administrative burden, but most remain cautious about deploying AI in high-risk clinical scenarios, a Dec. 29 KLAS Research report found.
The report is based on interviews and surveys with 3,370 respondents from 1,742 healthcare organizations, including provider and payer organizations, conducted throughout 2025. It is a perception report that identifies which AI use cases and vendors are most commonly used or considered by organizations. It does not evaluate vendor performance or market share.
Here are eight key findings from the report:
- Most organizations are piloting or using AI, but few have scaled it. Nearly every interviewed organization reported piloting or using some form of AI in 2025. However, few have expanded usage across departments or functions. Gating factors to broader adoption include lack of governance frameworks, the need for ROI validation and challenges integrating AI into existing workflows.
- Ambient speech is the most commonly adopted clinical AI tool. Seventy-nine percent of organizations reported using ambient speech technology to support clinical documentation, making it the most widely adopted clinical AI application. Respondents cited improvements in EHR documentation, while higher-risk tools, such as clinical decision support, remain limited to pilot phases.
- Organizations are focused on lower-risk use cases. AI investments are concentrated in well-defined workflows like documentation support, coding automation and administrative tasks. Use cases with potential patient impact are typically still in early stages due to safety and liability concerns.
- Generative AI is primarily used for administrative tasks. Twenty-two percent of organizations reported using generative AI, including tools like ChatGPT and Copilot, for meeting transcription and document creation. These tools are often deployed across multiple departments to streamline nonclinical workflows.
- Revenue cycle and imaging AI use cases are gaining traction. Revenue cycle applications — such as claims adjudication (24%), coding automation (24%) and denials management (17%) — are among the most commonly adopted. Imaging AI, particularly in neurology and breast imaging, is also well-established due to a longer history of use in healthcare.
- Microsoft and Epic are the top vendors in use or consideration. Microsoft was mentioned most often, largely for ambient speech and generative AI tools, followed by Epic, whose AI tools span multiple categories, including automated message responses, chart summarization and denials management. OpenAI, Abridge and several third-party ambient documentation vendors were also commonly cited.
- Despite significant industry attention on agentic AI, adoption is minimal. Of more than 3,000 respondents, only 17 specifically mentioned agentic AI, and just one organization reported current use. Most organizations said they are focused on deploying simpler AI tools first and are still building foundational data and governance infrastructure.
- Lack of governance frameworks, uncertainty around measurable ROI and difficulty integrating AI into existing systems were cited as the most common obstacles to enterprise-wide deployment. Many organizations are continuing to run targeted pilots to evaluate safety, oversight and value before expanding use cases.
