Over 90% of health systems have deployed third-party AI solutions, yet fewer than half have a dedicated data platform to test them. A new report from the Center for Connected Medicine at UPMC and KLAS Research reveals an industry racing to adopt AI in clinical and administrative workflows while the governance, strategy, and infrastructure needed to manage risk lag behind.
The findings, drawn from a survey of over two dozen health system leaders, show AI has moved firmly into mainstream use. But the operational scaffolding to validate performance, measure success, and support long-term value remains uneven at best.
Where AI is taking hold
Clinical documentation leads AI deployment areas, cited by 52% of respondents. Revenue cycle, coding, and billing applications follow close behind. The pattern points to an industry prioritizing tools that reduce administrative burden and address well-documented pain points in workflow efficiency.
Testing is nearly universal - 92% of organizations evaluate third-party AI tools before deployment. The methods, however, vary widely. Some rely on formal vendor testing protocols. Others run limited pilot programs or use informal internal reviews. Consistency across the sector is absent.
The infrastructure gap
Only 44% of health systems have a dedicated data platform or environment for testing AI solutions. That figure exposes a structural weakness. Without isolated testing environments, organizations struggle to evaluate safety, scalability, and real-world performance before tools reach clinicians and patients.
"The health care industry has moved remarkably quickly from discussing the potential of AI to actively deploying solutions across the enterprise," said UPMC Chief Medical Information Officer Rob Bart. "What's emerging from this research is a clear recognition that implementation is only the first step. Health systems are now focused on building the governance structures, testing capabilities and organizational strategies necessary to ensure AI delivers meaningful and measurable value."
Strategy and measurement still evolving
AI strategies at most organizations remain fluid. Sixty-three percent of respondents described their approach as developing or ad hoc rather than fully established or advanced. This tracks with the broader picture: deployment is happening faster than the planning that should underpin it.
No consensus exists on how to measure AI success. Respondents cited no shared metrics for evaluating return on investment, operational impact, or clinical value. The absence of standard benchmarks complicates decisions about scaling, renewing, or sunsetting AI investments.
Resource constraints compound the challenge. Health systems identified limited resources, insufficient time, and shortages of specialized talent as leading barriers to adoption. These are not novel problems in healthcare, but they hit harder when the technology demands rigorous, ongoing oversight.
Why this matters for executives and strategy
The report underscores a hard truth for health system leaders: adoption velocity does not equal organizational readiness. The gap between deploying AI and governing it creates exposure - to performance drift, compliance risk, and wasted spend. Executives who treat AI governance as a follow-on project rather than a prerequisite will find themselves retrofitting controls onto systems already in clinical and financial workflows. For strategy leaders, the data makes a clear case for prioritizing testing infrastructure and success metrics now, not after the next wave of vendor contracts is signed.
Your membership also unlocks:
