In 2026, health system IT leaders are focused on scaling artificial intelligence, moving from pilots and proofs of concept to operationalizing AI, but these efforts face significant barriers.
A new study from Qventus found that 74% of health system leaders cite dependence on electronic health record (EHR) vendors as a barrier to implementing their AI strategy.
More than 60 chief information officers, chief AI officers, chief medical information officers and other senior IT leadership members from health systems across the country were surveyed for the April 9 report. Leaders from these organizations, including HonorHealth and Rochester Regional Health, were asked how they are implementing AI tools and addressing reliance on EHRs.
“We think the trade-off for waiting on Epic and/or Oracle is that you could potentially lose money,” Matthew Anderson, MD, HonorHealth’s CMIO, said in the report. “Late movers have disadvantages. It’s not a first mover advantage, but late movers have disadvantages.”
However, the report found that surveyed organizations are less willing to put their EHR systems on hold to develop AI solutions, with only 22% of respondents saying they would wait for EHR functionality, down from 52% in 2025.
When asked whether health systems would rather wait 18 months for AI capabilities in their EHRs or implement a solution from a third-party vendor in three months with a guaranteed return on investment, the survey found that 40% said they would move to a third-party solution now, 38% said it would depend, and only 22% would wait.
AI tools have become commonplace in healthcare organizations across the country. Eliciting Insights’ March study found that 75% of U.S. health systems are using at least one AI platform, up from 59% in 2025. Additionally, 50% of respondents said their systems use three or more AI applications.
As the use and spending of these tools increases, so does the pressure to adopt them, with 65% of respondents in the Qventus survey ranking the pressure to operationalize AI at 7 or above on a scale of 1 to 10.
Measuring return on investment also remains a challenge for systems, with four out of five respondents reporting that measuring AI ROI is difficult. 39% report having no clear benchmarks for performance or ROI.
Additionally, 38% of respondents said they don’t expect it to take more than 13 months to see an ROI, while 74% said they need an ROI within a year to justify their AI investment.
Of the 60 CIOs, chief AI officers, CMIOs, and other senior leaders surveyed by Qventus, 42% reported actively deploying AI across multiple use cases, but only 4% achieved large-scale AI adoption with measurable outcomes. Nearly every quarter reports that pilot solutions are operational in limited regions.
“A wrong bet against technology can wipe out all your profits,” James Whitfill, M.D., chief transformation officer at HonorHealth, said in the report. “Not-for-profit health care makes grocery stores look like they’re profitable.”
Another barrier to the system is limited IT resources when managing multiple AI vendors, which 66% of respondents cited as an issue. 25% of respondents said they manage between 4 and 7 AI vendors. Nearly half of survey respondents report spending 11% to 25% of their IT bandwidth on vendor management, integration, and implementation alone, and 17% report spending as much as 26% to 50% of their organization’s IT bandwidth.
“Working with too many AI vendors increases system complexity and introduces additional unnecessary costs,” the report said. “This issue is not new. A few years ago, companies faced the proliferation of AI assistants, whose usage exploded across the enterprise and then had to be reined in.”
Despite the obstacles, 94% of respondents said delays in deploying AI would put their organization at a “competitive disadvantage” vis-à-vis other systems, and 68% of respondents said such delays would also increase clinician burnout and turnover.
“AI is certain to dramatically reshape healthcare systems in the coming years,” the report says. “But at what pace, at what price and at what scale are yet to be determined.”
