Healthcare’s AI story is changing. For several years, generative AI largely lived in the “pilot era,” marked by experimentation, point solutions and cautious optimism. In 2026, that era is giving way to something more operational: organizations are moving beyond testing AI tools and toward integrating them into care delivery.
Yaw Fellin, Senior Vice President and General Manager, Clinical Decision Support and Provider Solutions at Wolters Kluwer Health, sees healthcare in a transitional moment, where experimentation and operationalization must happen side by side.
“In some ways, I think we’ve absolutely moved into an operational phase,” Mr. Fellin said. “In other ways, I think continued experimentation is critical — we’re living in a little bit of a balanced phase.”
That dual approach, he said, is becoming essential as capabilities advance quickly and regulations and governance practices vary widely across organizations.
Becker’s Healthcare recently spoke with Mr. Fellin about the signals that mark this transition, exploring what meaningful expansion of AI looks like and why transparency and evidence-based support remain essential to clinician trust.
AI is moving deeper into systems’ operational fabric
From Mr. Fellin’s perspective, three trends indicate healthcare is moving beyond AI experimentation. The most visible, he said, is ambient documentation. While ambient tools began as pilots, adoption is now accelerating in a way that feels operational.
“The adoption of ambient continues, in my opinion, to really be at a much more operational level,” he said. “People have tried, they’ve piloted — they’re going full force.”
A second signal is the way AI is becoming embedded into healthcare technology itself. AI capabilities are increasingly built into core platforms, rather than existing as separate add-ons.
“If you just look at the number of products where AI capabilities have moved into the core product, that tells me we’re at a very operational phase,” he said. “It is becoming almost an expectation or a requirement.”
The third signal is governance. Organizations don’t build governance structures for one-off experiments, Mr. Fellin observed. They build them when something is becoming part of the operating environment.
“We’re seeing both tremendous need and a number of new governance processes appearing across hospitals and health systems,” Mr. Fellin said. “To me, that’s very much a signal that you’re at an operational phase.”
The new battleground for adoption: workflow integration
As AI becomes operational, workflow matters more than novelty. Leaders increasingly want tools that meet clinicians where they work.
“Workflow is basically how stuff gets done,” Mr. Fellin said. “When we talk about embedding into workflows, we have to ask if it’s participating in the place where people are doing what powers the business.”
At the same time, he believes the industry is heading into a larger question: Will the EHR remain the center of gravity for clinical workflows, or will AI-enabled experiences begin to shift that center?
“There’s actually a bigger macro question in the industry around how much will the core healthcare workflows move away from EHRs,” he said. “EHRs have dominated hospital and health system workflow for the last couple decades, and I don’t think anybody truly knows how that’s going to play out.”
Some believe ambient and AI-driven experiences could become the dominant workflow layer. Others believe these tools will become integrated into existing EHR workflows. Either way, Mr. Fellin said the key question remains the same: “Where are the actions happening?”
Clinical AI crosses a threshold
Not long ago, the prevailing view was that AI’s strongest early use cases were administrative, while clinical applications would move more slowly because of risk and complexity.
That is changing quickly.
Mr. Fellin pointed to Wolters Kluwer’s UpToDate Expert AI as one example of how quickly clinical AI can move into real-world use when the value proposition is clear and governance is in place. Though the product launched less than six months ago, in September 2025, more than 50 hospitals and health systems have fully deployed the AI-enabled tool for clinical decision support at the point of care.
Importantly, he said, these were not informal trials.
“All of those 50 went through a full governance process and reviews,” Mr. Fellin said. “This is not, ‘We’ve got a department or one or two docs adopting this.’ The organization has decided to adopt.”
Governance, however, remains a work in progress — and one that still varies greatly across organizations, Mr. Fellin noted. “For AI in particular, I would still say we’re in the early stages. Generally speaking, it’s more a set of bespoke processes,” he said.
He observed how some organizations are adapting existing playbooks, such as governance structures built for ambient documentation, and applying them to broader AI adoption. Others are building governance for the first time. Across the board, Mr. Fellin emphasized the importance of protocols and oversight that supports monitoring and responsible rollout, particularly in clinical decision support.
Defining ‘meaningful scale’ and demonstrating value
As more organizations look to expand AI across the enterprise, the term “scale” is everywhere. Mr. Fellin argues leaders should be careful about what they mean when they use it.
For him, scale implies more than simply expanding a tool to more users. It suggests a shift into sustainable adoption and enterprise value, which he breaks down into three core dimensions:
- Meaningful scale requires a cost-benefit balance. “I think cost-benefit ratio is really implied when people talk about scale and typically places where it’s still too costly for the return that you’re getting — that’s a big barrier to actually scaled use or scaled adoption of something,” Mr. Fellin said.
- Scale requires sustained use. Mr. Fellin said leaders should look for signs that usage holds and grows, rather than fading after the novelty wears off: “Once somebody starts using it, how likely are they to continue to use it? Generally, is the pattern of usage growing?”
- Scale implies extensibility. A scaled AI capability should enable other use cases, not remain limited to one narrow workflow. “It’s not just that the original use case becomes sustainable and has the right cost-benefit analysis — it’s that there are adjacent use cases that get unlocked beside it,” he said.
Where trust is won or lost: transparency at the point of care
Even as adoption grows, trust remains a defining barrier, especially in clinical decision support where stakes are high. For Mr. Fellin, trust is straightforward: “It basically boils down to: Can I rely on something?”
In clinical decision support, that reliance depends on accuracy, but also depth. Clinicians must be confident the information captures the full complexity of the decision at hand.
Transparency plays a central role in making that reliance possible.
“I view transparency as a building block of trust,” Mr. Fellin said. “The more transparent you can be, the easier it becomes for people to verify the answer and its basis, and to feel confident they can trust the response.”
Evidence-based medicine aligns closely with that expectation, he added, because it is rooted in rigorous study, peer review and scientific validation. When decisions carry risk, trust becomes non-negotiable.
Long-term value is more than efficiency
Time savings is often the first measurable benefit leaders point to, especially with ambient documentation. But Mr. Fellin believes AI’s most meaningful long-term value lies in improving people’s ability to generate outcomes — a goal he described as Wolters Kluwer’s north star.
“Speed and efficiency are super important,” he said. “But can you make patients healthier, impact the community, lower costs? A lot of these technologies, especially the expansion from administrative to clinical use cases for AI, I absolutely believe have that potential.”
For organizations earlier in their AI journey, Mr. Fellin recommends starting with disciplined experimentation: defining a clear hypothesis up front and establishing how success will be measured. From there, leaders can expand what works and build toward operational adoption using structured approaches such as lean process techniques.
As healthcare scales AI, Mr. Fellin urged leaders not to lose sight of the human component, even as excitement builds around new capabilities. In clinical decision support, he believes expertise must remain embedded in AI-enabled workflows.
“I still don’t believe that a human in the loop is enough,” he said. “You actually need experts in the loop of these AI systems.”
In a landscape moving from pilots to enterprise wide deployment, that may be the central leadership challenge: scaling AI with ambition, while keeping trust, governance and clinical expertise at the center.
At the Becker's 11th Annual IT + Revenue Cycle Conference: The Future of AI & Digital Health, taking place September 14–17 in Chicago, healthcare executives and digital leaders from across the country will come together to explore how AI, interoperability, cybersecurity, and revenue cycle innovation are transforming care delivery, strengthening financial performance, and driving the next era of digital health. Apply for complimentary registration now.
