The Technologies That Actually Help: Why Healthcare Innovation Should Amplify, Not Replace, Human Expertise
I recently found myself in a conversation with a colleague who splits time between emergency medicine and healthcare informatics—a combination that offers a uniquely sobering perspective on our industry's endless parade of technological solutions. What struck me wasn't their skepticism about AI in medicine, but rather their nuanced view of what actually works—and what doesn't—in healthcare innovation.
"We've definitely hit a critical point when two of the hospitals in my health system each have their own generative AI application with different outcomes for making perfect notes," they told me. "Do we need another ambient AI scribe?"
But this wasn't dismissive cynicism. In fact, my colleague had enthusiastically implemented scribing tools and received "love letters from physician colleagues that are like you've changed my life."
The Technology That Actually Works
This distinction reveals something crucial about healthcare innovation that most discussions miss: the difference between solving human problems and solving technical ones. Ambient scribing works because it addresses a genuine pain point that directly affects patient care—the administrative burden that pulls physicians away from their patients.
It's one of the few AI applications that actually enhances the human experience of medicine rather than trying to replace human judgment.
The conversation revealed a more sophisticated framework for thinking about healthcare technology. Not all innovation is created equal, and the most successful interventions tend to be those that amplify human capabilities rather than automate them away. "I actually think scribing and getting rid of tedium is great," my colleague explained, contrasting this with other AI applications that promise to revolutionize clinical decision-making.
The False Promise of Prediction
Consider the difference between ambient scribing and sepsis prediction algorithms. The scribing technology succeeds because it tackles a clear, defined problem: documentation burden interferes with patient interaction and contributes to physician burnout. The solution is straightforward—capture the conversation and generate the note—and the benefit is immediate and measurable.
Sepsis algorithms, by contrast, attempt to solve a much more complex problem: clinical uncertainty in diagnosis. But as my colleague noted, "There's always some caveat or catch where it's like well we can predict sepsis if you order these 14 tests and then we'll predict sepsis."
The logical outcome? Everyone orders everything upfront. The algorithm doesn't reduce uncertainty; it systematizes defensive medicine.
This speaks to a deeper insight about what makes healthcare technology successful. The most effective innovations don't try to replace clinical judgment—they remove barriers to exercising it well. Scribing technology exemplifies this principle: it doesn't tell doctors what to think, but it frees them to think clearly by eliminating the cognitive overhead of documentation.
The Reality of Healthcare Complexity
The broader pattern my colleague identified is that healthcare's most intractable problems aren't technical—they're structural and human. "Most people are in a messy, unique situation that isn't AIable," they observed.
The fantasy underlying much healthcare innovation is that complex, interconnected human problems can be solved through discrete technological interventions.
But the success of ambient scribing suggests a different path forward. Instead of building AI systems that attempt to make clinical decisions, perhaps we should focus on AI that makes it easier for humans to make good decisions. Instead of predicting sepsis, maybe we should be building tools that help physicians communicate more effectively with patients about uncertainty and risk.
The Market Reality Check
The switching costs for many AI tools, my colleague noted, are essentially nothing. Physicians in their system regularly experiment with different ambient scribing platforms, choosing tools that work best for their practice patterns. "The switching cost is nothing. It takes literally nothing to spin up" these applications.
This commoditization might seem threatening to AI companies, but it's actually a sign of a healthy market—one where value is determined by actual utility rather than vendor lock-in.
What's remarkable about the scribing success story is how it emerged organically from physician needs rather than being imposed by administrators or technologists. My colleague described implementing these tools because physicians were asking for them, not because some innovation committee had identified documentation as a strategic priority.
A Different Path Forward
This bottom-up adoption pattern suggests that the most transformative healthcare technologies might be the least ambitious ones—tools that solve specific, well-defined problems rather than promising to revolutionize medicine.
The irony is that by aiming lower, these technologies might actually achieve more meaningful change.
The conversation left me wondering whether our obsession with breakthrough innovation might be preventing us from recognizing incremental improvements that genuinely matter. Maybe what healthcare needs isn't disruption, but better tools that help humans do what they already do well.
Sometimes the most profound innovation is simply getting the mundane stuff out of the way so that human expertise can flourish.
As my colleague put it: "I enjoy being in the middle of the flaming dumpster fire rather than any one thing I want to pin my hat on for the rest of my career." There's wisdom in that embrace of complexity, and perhaps the best technology is that which supports rather than supplants that very human commitment to wrestling with difficult problems.