There is a graveyard in healthcare AI that never appears in health system announcements or investor decks. It is filled with tools that earned high accuracy scores in validation studies, wowed executives in demonstrations, and generated genuine excitement in pilots — before quietly dying when they met the reality of clinical workflow.
The leaders responsible for deploying these tools have seen the pattern repeat often enough that they’ve given up blaming the technology. The model is almost never the problem. The problem is everything around it.
Thomas Kingsley, Director of Applied AI for UCLA Health, has watched this failure mode play out with enough regularity to describe it precisely. The culprit isn’t a flawed algorithm. It’s a flawed theory of adoption.
“Most promising AI tools fail not because they don’t work, but because they weren’t designed with the clinical workflow, user experience and implementation context in mind,” he said. “A perfectly accurate model that adds 30 seconds to a physician’s workflow per patient will quietly die on the vine. The hard problems in healthcare AI are sociotechnical – governance, trust, integration and sustained change management – not algorithmic.”
In busy clinical environments, even marginal friction compounds into a significant burden with many patient encounters. Tools don’t always account for that reality and clinicians simply stop using them. The tools with the best adoption rates aren’t the flashiest tools, but instead the ones fixing day-to-day drudgery.
The counterintuitive implication is that the most successful AI deployments will be the ones users barely notice. The technology recedes into the background, reducing friction rather than adding a new layer of interaction.
Sunil Dadlani, executive vice president and chief information and digital transformation officer for Morristown, N.J.-based Atlantic Health, has seen this dynamic play out in the Epic ecosystem, where integration into an existing clinical environment consistently outperforms standalone platforms, regardless of the latter’s technical sophistication.
“There is a common assumption that the most advanced or largest model will automatically drive the strongest clinical or operational results. In healthcare, that is rarely true,” he said. “Context and workflow integration matter far more than model size. A focused, well-tuned model embedded directly into Epic workflows often outperforms a highly sophisticated standalone platform that sits outside of the clinician’s daily environment.”
Similarly, Hetal Rupani, senior director of business intelligence and analytics for Baltimore-based Johns Hopkins Medicine, told Becker’s that having a highly accurate, “perfect” model doesn’t guarantee high impact.
“Real value depends far more on how effectively AI is integrated into clinical workflows, EHR systems and aligned across strategy, data and people,” he said. “Efficiency gains only materialize when AI meaningfully improves how work gets done – not when it simply performs well in isolation.”
Identifying the right tools to integrate into clinical workflows and giving strong feedback–or co-developing applications or platforms, takes time but is worth the effort. The tools need to be easy, trustworthy and show value to clinicians right away–not just the bottom line–to secure adoption. There are downstream consequences of skipping the human side of technology implementation.
“I’ve seen near-perfect algorithms collect dust because no one asked the clinicians first,” said Jennifer Ngure, director of clinical informatics at Attleboro, Ma.-based Sturdy Health. “The hospitals winning on AI aren’t those with the best tools–they’re the ones who treated implementation as a human problem before a technical one.”
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.
