Many providers express confidence about their AI strategy, but relatively few have established the governance structures needed to ensure responsible AI deployment, according to new research.
Nordic Consulting released a report this month based on a survey of 127 leaders who work at healthcare organizations, mainly hospitals and clinics. The results showed that 70% of leaders feel at least somewhat confident in their organization’s AI governance frameworks, but only 15% report having scalable infrastructure in place.
While there’s significant enthusiasm about AI, scaling it across a healthcare enterprise is proving to be an incredibly complex process, said Kevin Erdal, senior vice president of transformation and innovation services at Nordic.
To achieve scalability, providers have to take a deep dive into what “scale” really means in terms of sustained use. Many organizations underestimate the ongoing management needs of AI models, especially custom tools that consume high computing resources, Erdal stated.
Data readiness is also crucial for AI success, he pointed out. Many survey respondents cited the lack of infrastructure to access and process data from across disparate systems as a major barrier to AI scalability, Erdal noted.
“It might be a scenario in which you already have the data readily available or stored, but you don’t necessarily have the interoperability to reach out and grab some of the data from your pertinent systems across the collective institution. It’s one thing to be able to store data, and it’s another to be able to process data,” he explained.
When it comes to new AI tools on the healthcare market, there’s plenty of hype and flashiness — but it’s the unsexy, foundational elements, like data management and computing infrastructure, that determine real victories, Erdal declared.
If organizations can’t capture the right data, models will fail — no matter how promising the technology, he warned.
He also pointed out that healthcare leaders may overestimate their AI readiness due to the wide availability of vendor models out there. In his eyes, true preparation includes governance, infrastructure, data and — critically — change management.
“It’s one thing to turn a model on, but do you need the overarching governance to bring in those operational users into that conversation,” Erdal remarked.
The change management process is often overlooked, and organizations often fail to explain the overall goal of the technology to its end users, he explained. For instance, if a hospital deploys an AI model to predict no-shows, the organization must communicate its plan for what to do with that insight, Erdal said.
As AI adoption continues to evolve in healthcare, success won’t come from flashy demos — It will come from the less glamorous work like this, he declared.
Photo: erhui1979, Getty Images
