Balancing health AI management with growing vendor sprawl

Multiple AI vendors fill gaps, meet specialist needs

Though core systems, especially the EHR, remain central to health IT operations, health systems are turning to multiple AI vendors to address needs that a single platform cannot fully support.

At the University of Arkansas for Medical Sciences, for instance, a variety of AI tools have been deployed across nearly every step of the patient journey, from intake to follow-up, Joseph Sanford, M.D., associate vice chancellor and chief clinical informatics officer, said.

That approach is common across the industry. The survey data shows 25% of health systems are currently managing between four and seven AI vendors. Providers say that the level of diversification reflects an operational reality rather than fragmentation.

"One vendor just does not make sense," said Matthew Anderson, M.D., chief medical information officer at HonorHealth.

Even large EHR vendors cannot cover every use case across specialties and workflows.

"Epic cannot provide every single solution for every specialty in every location," Anderson said.

Sanford echoed that limitation, noting that EHR platforms are foundational but incomplete.

"Epic…cannot be everywhere all at once," he said.

That gap has created space for additional tools designed to address specific clinical and operational challenges.

"There is, I think, an open opportunity for meaningful market entry for point solutions that have a competitive advantage and are solving for a niche that is important, but perhaps under-resourced," Sanford said. 

EHRs guide AI strategy

EHR systems are central to AI strategy for many health systems, serving as the foundation for data, workflows and core clinical operations. Sanford noted that because EHR systems represent a major investment, organizations are expected to fully leverage those capabilities before pursuing additional AI tools.

"You have to start with what you have already paid for," he said.

But reliance on EHR vendors can create its own constraints. Survey data shows 74% of health systems say staying dependent on their EHR vendor's AI roadmap -- the timeline and priorities for when new AI features are developed and released -- is holding them back. Organizations are often forced to weigh whether to wait for new capabilities or pursue external solutions.

"If something is three years away and I have a six-month problem, you are going to have a conversation [about other options]," Sanford said.

Although survey data shows that 72% of health systems would prefer a single, streamlined AI partner, having options is critical to maintain flexibility.

"I think there is a strength in optionality," Sanford said.

The challenge is how those tools are managed in practice. Health systems are working to integrate a growing number of systems into clinical workflows, manage data across vendors and govern their use in real time.

"Every new vendor…is creating another silo, and it is another silo that can fail," Anderson said.

Managing the AI burden across teams

Managing a growing ecosystem of AI tools requires coordination across clinical, IT and operational teams, often stretching already limited resources. Survey data shows 51% of health systems spend 11% to 25% of IT bandwidth on vendor management, integration and implementation. Some reported they spend as much as half of their capacity on those efforts.

At HonorHealth, that work is distributed across a broad group of stakeholders rather than centralized in a single team.

"It takes a village…a village of street-smart, security-alert people," said Anderson.

Still, those responsibilities -- including vendor evaluation, implementation, data governance and oversight -- are often layered onto existing roles, adding to the operational burden.

"We have a pretty big team of people who all have other jobs…who are now working on evaluating and implementing AI," Anderson said.

At the University of Arkansas, leaders describe a similar challenge as they work to prioritize and govern a rapidly expanding set of AI opportunities.

"There are far more potential opportunities than any organization has the capacity to handle at the same time," said Sanford.

He also noted that the pace of change has made governance an ongoing process rather than a fixed framework.

"The moment the ink has dried on your governance policy, it is outdated, and you need to start over," he said. 

Ongoing fragmentation, workflow and data challenges

Integrating multiple AI tools into clinical workflows remains a persistent challenge, often limiting the impact of the otherwise promising technologies. Survey data shows 45% of health systems cited challenges scaling AI pilots into production.

That gap often emerges when tools that perform well in isolation fail to fit into the broader care process. Anderson noted that improving one step does not necessarily improve the whole workflow.

"If it makes the preceding step or the next step worse, you are not really helping," said Anderson.

Fragmentation across systems can introduce inefficiencies, increase cognitive burden and create new opportunities for error. Anderson said that seamless integration of AI tools is no longer just nice to have at the organization; it's a necessity

"Logging into one more thing…cut and pasting from one thing to another, it just does not work," he said.

Data governance adds another layer of complexity, particularly as organizations evaluate how information is shared across vendors.

"I think the data is the key point…what happens with that data? Where does it go? Who owns it?" Anderson said.

As organizations push to move beyond pilots and demonstrate value, the ability to integrate and manage AI systems is becoming more important than adding new ones. Until integration and management are smoothed, adding more AI tools could slow progress rather than accelerate it.

Elizabeth Stricker, BSN, RN, comes from a nursing and healthcare leadership background, and covers health technology and leadership trends for B2B audiences.