AI in healthcare: From wide adoption to operational accountability

Healthcare organizations have moved past artificial intelligence experimentation. In RSM’s 2026 AI survey of middle market organizations, 84% of healthcare respondents whose organization uses AI reported that AI is already partially or fully integrated across operations and processes. To bring those numbers to life, RSM conducted two focus group conversations* with healthcare executives to gain further perspectives on AI uses, challenges and more. What follows are the key themes and insights that emerged from the survey and focus groups.

AI is widely adopted, unevenly operationalized

Of the 70 health care respondents who participated, 40% reported full AI integration across core processes, and another 44% reported partial integration. Inside organizations, the picture is more layered. Leaders described AI as often leveraged through embedded vendor tools rather than implemented as a deliberate strategy, with maturity varying widely across functions.

Administrative use cases are scaling far faster than clinical ones. As one focus group participant put it: “From an administrative perspective, very high. From a clinical perspective, probably not 84%.” Ambient documentation has emerged as the clearest near-term clinical bright spot, described in the same discussion as “growing like wildfire with our provider community.”

AI governance exists, but maturity varies

The survey shows AI strategy ownership is distributed across the CEO (27%), IT (26%) and the senior AI roles such as the chief AI officer and the head of AI or AI center of excellence (24%). And 94% of healthcare respondents reported having governance controls in place, yet 13% said controls are applied only after issues arise, and 9% said they are applied inconsistently.

Focus group participants acknowledged adoption frequently outpaces oversight. “People start using these tools a little bit faster than sometimes we would like until you have governance,” one participant said.

Another captured the challenge bluntly: “We’re trying to be compliant without knowing what compliance is.”

Governance that builds confidence is designed early on, not retrofitted.

Measuring AI ROI: Optimism tempered by complexity

Healthcare organizations reported they evaluate AI return on investment most often through: 

Optimism is high: Among those who measure AI ROI (99%) a large majority said AI has met or exceeded their expectations for ROI (90%). That optimism comes with caveats. Focus group members described persistent difficulty translating soft gains into board-level outcomes.

“Everybody comes out, says you’re going to become more efficient,” one participant noted. “But when I say ‘Let’s quantify that,’ it’s crickets.”

AI’s most defensible value today is its ability to increase financial and operational efficiency and address clinical and capacity needs, though real return on investment dollars remains harder to quantify.

Defining AI success

When asked what would signal AI success in the coming year, healthcare survey respondents most commonly cited:

Focus group participants framed success differently, moving away from one-off tools and toward reusable platforms, fewer experiments and more confidence in governance.

“A year from now, what I don’t want to be is talking about AI,” one healthcare executive said. “What I want to talk about is solutions—what problems has it solved?”

Weighing AI investment and strategy

AI usage decisions increasingly involve prioritization, not just enthusiasm. The survey found that 86% of healthcare respondents expect their workforce size or composition to look fundamentally different in two to three years because of AI, and 79% expect AI spend to increase next fiscal year.

Among organizations expecting an increase AI spend, they most commonly expect reductions in:

Most organizations are pursuing or plan to pursue a hybrid AI model—building additional capabilities in-house (89%), buying from external providers (86%) and engaging external advisors (84%)—to balance speed with capability.

*Focus groups: In May 2026, RSM conducted two one-hour virtual focus groups with nine senior-level IT and technology professionals recruited through the College of Healthcare Information Management Executives (CHIME) and the Healthcare Information and Management Systems Society (HIMSS). All participants were based in the United States, with five holding a membership in CHIME and four holding a membership in HIMSS. Results from the focus groups are qualitative in nature and cannot be projected to the full population.