The healthcare industry is deploying AI faster than it can govern it, UPMC report finds
90.5 WESA |
By
Kiley Koscinski
Published August 11, 2026 at 5:31 AM EDT
Gene J. Puskar
/
AP
Most health systems have moved beyond experimenting with artificial intelligence toward an era of relying on the technology to manage tasks like clinical documentation and billing. But a new report from UPMC finds that, despite the widespread adoption of AI, many health systems are operating without clear guardrails to manage risks
“I think we all see the promise in AI, but [systems] need to have true guardrails … to actually implement those [tools] in an effective way across the health system,” said Ken Howard, vice president of UPMC Enterprises, the innovation and venture capital arm of the UPMC health system
The report, from UPMC’s Center for Connected Medicine and KLAS Research, surveyed more than two dozen healthcare leaders, including health system executives and ambulatory care organizations
The findings depict an industry eagerly adopting AI tools. But when it comes to the infrastructure surrounding how to apply those tools without risk or safety concern, many health systems appear to be building the plane while flying it
Among the most cited uses of AI were clinical documentation, imaging, revenue planning, patient engagement, workforce tools and productivity assistants
The report found that while most (93%) health systems have deployed third-party AI tools, a majority (63%) of those surveyed are still developing the infrastructure to support them, describing policies as “ad hoc” rather than established or advanced
Even fewer respondents (44%) reported having a dedicated data platform to test AI solutions using models adapted from real-world patients. Still, 92% of respondents reported testing third-party tools prior to deployment through pilot test or proof-of-concept work from the third-party vendor
Without good internal infrastructure, the report contends, systems may struggle to validate the findings of their AI tools. Respondents said data quality can vary widely based on inconsistent definitions across different teams, the use of manual workarounds and spreadsheets of unstructured data. The report suggests that many systems are still relying on labor-intensive processes to reconcile and prepare data for reporting or for use by an AI tool
UPMC has deployed its own data platform called Ahavi to test AI applications without disrupting patient care. Howard said the platform uses de-identified patient data to mimic real-world scenarios
“As AI adoption continues to expand, healthcare providers are recognizing that success depends on more than selecting the right technology,” Howard said
But what makes for successful use of AI in the health setting is also a point of contention, the report found. There was a lack of consensus from health systems about what a successful AI strategy looks like as well as no universal metric to determine the value the tech brings to patients or system efficiency
The report suggests that AI challenges in health systems go beyond technical snafus and into operational and procedural. And those challenges continue even after an AI tool is deployed
UPMC contends that ongoing governance, testing and evaluation is critical to deploy AI responsibly in the healthcare setting
“What’s emerging from this research is a clear recognition that implementation is only the first step,” said Rob Bart, M.D., chief medical information officer at UPMC. “Health systems are now focused on building the governance structures, testing capabilities and organizational strategies necessary to ensure AI delivers meaningful and measurable value.”
