AI is now embedded in everyday clinical work, helping physicians and nurses retrieve information faster and offload documentation. Recent Wolters Kluwer research found that 74% of physicians and 70% of nurses use generative AI weekly — yet 92% still say human expert validation is essential.
That gap, between fast adoption and an enduring demand for human judgment, is where health system leaders now have to operate: keeping clinical expertise sharp rather than letting it recede as the technology takes on more of the cognitive load.
During a Becker’s Healthcare webinar sponsored by Wolters Kluwer, five leaders discussed how AI is reshaping clinical reasoning and patient expectations, and what it takes to build organizations where AI strengthens clinical expertise instead of eroding it. The panelists were:
- Lisa Ivanjack, MD, chief medical officer at Providence Medical Group (Renton, Wash.)
- John-Paul Mead, MD, CMIO at Centralus Health (Ithaca, N.Y.)
- William Gustin, MD, chief medical officer and CMIO at Moab (Utah) Regional Hospital
- Amanda Heidemann, MD, senior clinical content consultant, clinical effectiveness, at Wolters Kluwer Health
- Yaw Fellin, senior vice president and general manager, clinical decision support and provider solutions, at Wolters Kluwer Health
Here are four key takeaways from the discussion.
Note: Quotes have been edited lightly for length and clarity.
1. AI scribe adoption was fast. Scaling beyond it is the real test.
Ambient AI scribes have become table stakes, panelists agreed, and adoption has moved quickly across care settings. The harder work is what comes next.
“Our providers said, ‘I’ll take an inch and then I want a mile. So what else can it do?'” said Dr. Gustin, whose critical access hospital has used AI for a couple of years. He described rapid uptake across the emergency department and inpatient admission and discharge, but cautioned that expanding past the scribe asks clinicians to rebuild ingrained habits.
“This idea of relearning documentation, ordering, workflow, it’s a fair amount of burden just to rethink your muscle memory to include it,” Dr. Gustin said.
Dr. Ivanjack described AI scribes as “ubiquitous” in most health systems but noted adoption has lagged among hospital-based physicians and specialists, even as primary care has embraced the tools — leaving “room for improvement on both the technology side, as well as on implementation.” For clinicians already using ambient documentation, she said, surveys show it has shaved hours a week off work and after-hours “pajama time.”
2. The goal is cognitive balance, not just less cognitive load.
Several panelists pushed back on the idea that AI’s value is simply removing work. Used well, they argued, it can redistribute a clinician’s mental effort toward the judgment that only a clinician can supply.
Dr. Gustin framed the tool for his physicians as “the smartest intern you’ll ever have,” and to “trust but verify” — warning about the risk of anchoring bias if a clinician accepts an AI’s first impression without weighing other possibilities.
Dr. Mead drew a related distinction from resident training, cautioning that AI should sharpen, not shortcut, reasoning. Clinicians, he said, have to know they are “using it to broaden your differential” rather than defaulting to “the convenience bias of the AI.”
Mr. Fellin from Wolters Kluwer picked up a phrase from a customer to describe what good design should do: build “cognitive scaffolding” into the system that reinforces how a clinician reasons rather than doing the reasoning for them. “We’re walking people through the steps of a differential, that type of scaffolding,” Mr. Fellin said, tying the approach to the goal he cares most about — using technology at the point of care to drive better outcomes.
3. Trust depends on knowing where the information comes from.
The panel returned repeatedly to provenance and clinical intelligence: clinicians and patients can only trust AI outputs they can trace to a credible source. Dr. Mead said his system steers clinicians toward a “known corpus of knowledge” — trusted references and journals — rather than the open internet, and keeps EHR-embedded AI limited to drafting that a clinician must read and own before signing.
Mr. Fellin said evidence-based grounding is central to how Wolters Kluwer engineers its AI, describing a rigorous editorial process behind its guidance and the particular difficulty of getting high-stakes answers right. Drug dosing, he said, was the hardest problem the company faced.
“You do not want the AI to answer with the wrong dosage because that’s something that has a real potential for patient impact and patient harm,” Mr. Fellin said. Rather than decline those questions outright, the company built a system to measure the reliability of each response — and is now experimenting with showing users where a model’s own reasoning, versus sourced evidence, shaped an answer.
Dr. Ivanjack tied provenance directly to patient trust: knowing where information comes from is what lets a clinician validate the reasoning and reassure patients that the human is still making the decision, with AI as support.
4. AI literacy extends to patients, not just clinicians.
AI literacy came up across the panel as essential, but the speakers stressed that clinicians and patients start from very different places. Mr. Fellin flagged a dimension many organizations have not yet addressed: patients are already using AI extensively and bringing it into the exam room. “What does AI literacy look like, not just for clinicians, but also for patients?” he asked, suggesting health systems may have a role in guiding how patients use these tools.
For clinicians, panelists pointed to transparency and consent as the foundation. Dr. Ivanjack said patients need to know when and where AI is used and that it is not “replacing or usurping their relationship with their clinicians,” while noting AI can help translate care instructions into plain language patients actually understand. Dr. Gustin described tackling “AI fear” before literacy — showing hesitant patients the scribe, reassuring them they stay in control, and offering a documented opt-out for the few who decline.
Dr. Heidemann argued that the most useful literacy tools actively teach as they work, surfacing “nudges” that prompt a learner to consider what else might be driving a recommendation so clinicians build durable reasoning skills for the moments when no AI is at hand.
What separates the organizations that get this right
When asked what will distinguish health systems that use AI to empower clinicians, the panel converged on governance, paired with a willingness to move. Dr. Ivanjack said the systems that succeed are “neither rapid early adopters without any governance or avoiders,” but organizations that adopt wisely with attention to safety, data integrity and transparency.
Dr. Mead detailed the machinery that requires: a promulgated policy, a committee charter, a risk-leveling tool for vendor claims, and executive sponsors accountable to KPIs. Dr. Gustin urged leaders to stay nimble and experiment, imagining workflows several years out and working backward. And Dr. Heidemann offered a reframe for skeptics, recalling the EHR rollout and the push to have clinicians work to the top of their license.
“I think when we reframe that, ‘You’ve done this before, you’ve survived this transition before. It’s okay, these are the skills you have that you’ve used before.’ I think people become a little more trusting, and it’s a more productive interaction,” Dr. Heidemann said.
The through-line, as Mr. Fellin put it, is trust and transparency directed at what each organization is actually there to do — “advancing the mission of each of the organizations in patient care and outcomes.”
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.
