Headlines about AI in healthcare often promise a revolution. But as KLAS has interviewed healthcare leaders over the course of 2025, what we have actually seen is something more grounded. There is no doubt that AI is being embedded in real workflows, but instead of reinventing care overnight, organizations are using AI to shore up strained operations, protect margins, and make clinicians’ lives more sustainable.
Our recently published Healthcare AI Update 2025 report builds on the first Healthcare AI report from earlier this year, providing a nearly year-long view of what healthcare organizations are actually doing with AI, where adoption is growing, and where the hype still outpaces reality.
AI Sees Increasing Adoption but Mostly for Well-Defined, Lower-Risk Use Cases
Throughout 2025, the number of organizations reporting they use AI climbed from under half of respondents to more than two-thirds; users include different-sized payers and acute, ambulatory, and post–acute care organizations. Executives use AI the most, but the fastest growth is among directors and managers: an early sign that more-established AI use cases are expanding throughout organizations.
Even so, most organizations are not trying to transform care delivery in one leap. They are deploying AI where the benefits are clear and the risk is low: tasks like documentation support, meeting and call transcription, administrative work, imaging triage, and revenue-cycle automation. Ambient speech, administrative/operational tasks, imaging, and revenue cycle management make up the bulk of current use cases.
As we interviewed healthcare leaders, a consistent theme emerged: Before they trust AI in higher-stakes clinical decisions, they want stronger governance, clean data, and demonstrable ROI. Many organizations are still building the committees, policies, and guardrails they need to manage risk. That deliberate pace is why, despite the hype, AI’s impact so far is less about dramatic reinvention and more about doing the same essential work with fewer manual steps and fewer exhausted staff members.
Ambient Speech Has Become Healthcare’s First Big AI Success Story
If there is one AI use case that truly dominates the market, it’s ambient speech. Tools that listen to patient-provider conversations and generate structured notes have quickly become the standard entry point for many organizations’ AI journeys. In the report, ambient speech stands out as the most commonly cited clinical AI use case by a wide margin.
This is because clinicians are under enormous documentation pressure; anything that helps them finish notes the same day, capture more complete information, and spend more of the visit actually looking at patients instead of screens feels like a life raft. Leaders also see ambient speech as lower risk, as it augments human judgment rather than making independent clinical decisions.
Of note, more EHR vendors are releasing their own ambient capabilities within their core clinical platforms. That is nudging some organizations to weigh tighter integration and consolidated systems against niche tools that may move faster on innovation. We see this tension as a sign of maturation—ambient speech is no longer an experimental add-on; it’s becoming core infrastructure.
Revenue Cycle and Patient Engagement Are Emerging as the Next AI Focus
Organizations are increasingly expanding into revenue cycle use cases; they are experimenting with AI for coding support, claims adjudication, prior authorization, denial prevention, and payment estimation. These use cases are attractive because ROI is tangible—faster reimbursement, fewer errors, and more complete capture of money already earned.
Patient/member engagement is following a similar arc. Many organizations are using AI to automate responses to portal messages, power chatbots, and route or handle calls in their contact centers. Leaders tell us they are looking for ways to scale personalized communication without endlessly adding FTEs. When done well, AI can answer common questions, help patients navigate care, and free staff to focus on the more complex, human-intensive interactions that truly need them.
Agentic AI Is Still in Its Early Days
Given the extent to which vendors advertise agentic AI, the number of interviewed organizations who mention agentic AI is well below the hype. Across thousands of reported AI use cases, explicit mentions of agentic AI are a small fraction.
Because agentic AI solutions often depend on additional data sources and involve greater implementation complexity, they present an even higher barrier to entry. As a result, organizations focused on quick, demonstrable ROI have been slow to adopt these more advanced tools without clearer proof points of their value.
Reported use cases for agentic AI are diverse, but the most common use case is the automation of patient interactions (e.g., scheduling appointments, closing care gaps, gathering patient information). The majority of respondents haven’t yet identified specific vendors to consider.
Where We Go from Here
Looking ahead, we expect 2026 to be less about chasing every shiny AI object and more about stabilizing with AI; organizations will likely consolidate around AI use cases that actually deliver value and selectively pilot more advanced approaches where they have clear guardrails and a path to ROI.
As we continue tracking this space, our goal is to help healthcare organizations see what’s real, understand where peers are finding value, and avoid feeling either falsely behind or prematurely done with AI. For a deeper look at the specific use cases gaining traction, how adoption shifted over the course of 2025, and which solution categories are drawing the most consideration, read the Healthcare AI and Healthcare AI Update reports.
© Gorodenkoff / Adobe Stock
