Is the healthcare sector experiencing AI frenzy? Emphatically, yes. If you were around for the dot-com boom (and bust), today’s AI surge in healthcare should feel uncomfortably familiar. We’re seeing the same rush to invest capital and urgency to adopt AI well before the evidence demonstrates consistent, reliable, and repeatable results. This moment, however, feels different. And in some important ways, it is.
The data tells us this is not just hype layered on top of healthcare. A growing percentage of hospitals already have generative AI embedded in their EHR workflows. Physician adoption has surged, more than doubling in just a few years. More than half of all digital health investment is now flowing into AI-enabled companies. The combination of provider adoption, clinician usage, and capital concentration has created a sense of inevitability.
In almost every board discussion, vendor pitch and webinar, every strategy session seems to come back to the same question: “What are we doing about AI?” Rather than asking “Should we?”, we’re often asking “How fast can we?”
Frenzy is rarely about just one thing. It is usually the convergence of multiple forces which is exactly what we are seeing.
First, the economic pressure is real. Margins are tight. Workforce shortages persist. Administrative burden remains one of the most cited drivers of burnout. When any technology promises to reduce documentation time, improve revenue cycle performance, or expand access without adding staff, it understandably gets immediate attention.
Second, for the first time, we are seeing visible, near-term wins. Ambient documentation is a clear example. Organizations are reporting meaningful reductions in physician burden and improvements in patient interaction. When clinicians begin to say, “this actually helps me,” adoption accelerates quickly.
Third, because AI is being embedded into the core platforms and 3rd party applications that run healthcare, it is no longer a standalone experiment. When capabilities show up inside the EHR, adoption shifts from optional to inevitable. That raises the all-important question, how can we change our focus to ensure we’re achieving real transformation not just speeding up existing processes? Is the technology reliable enough to use in clinical and non-clinical operations? Do we understand and trust the algorithms that are associated with AI?
Fourth, the capital markets and vendor ecosystem are amplifying everything. When more than half of investment dollars are chasing AI, every company suddenly has an AI story to share. Some of those stories are real. Some are not. But all of them contribute to the excitement, noise, and pressure.
And finally, there is a growing sense of strategic anxiety. Health systems are not just asking whether AI can help. They are asking what happens if they fall behind (aka FOMO, the fear of missing out), one of the most powerful accelerants of all.
So yes, this is a frenzy. But it is a rational one. The challenge is that even rational frenzies still create risk. We are already seeing warning signs. Physicians remain concerned about privacy, safety, and loss of skills. Governance models are struggling to keep pace with deployment. And in some cases, organizations are implementing faster than they can evaluate. The rigor of clinical trials is rarely applied. We rush to implement.
This is where leadership matters. The answer is not to slow down indiscriminately, nor is it to accelerate blindly. The answer is disciplined acceleration.
Language influences behavior. As people have rushed to embrace innovation particularly in AI, I hear all too often the phrase, “fail fast” instead of “succeed quickly”. Failing fast centers on failure, not outcomes. It allows, even encourages poor discipline, and creates cultural ambiguity. “Succeed quickly” is outcome oriented and requires thoughtful design and building. It is more aligned with executive thinking and reinforces a culture that rewards discipline, focus, and accountability. Ultimately, it encourages practical and pragmatic innovation.
Where should we start? Start where the value is clearest and the risk is lowest, e.g., administrative workflows, documentation, patient access, and revenue cycle. These are areas where the evidence for near-term return is strongest and patient harm is less likely. These kinds of incremental successes build confidence and can lead to long-term sustainability. Success will create even more demand. As a CIO, I often reminded my staff to use our governance processes to manage demand and when asked to do more, to diplomatically respond, “We can do anything, we just can’t do everything.”
As risk increases, apply increasing levels of rigor. Clinical decision support, patient-facing AI, and autonomous workflows demand stronger validation, monitoring, and governance, like clinical trials for medications. Treat AI as an enterprise capability, not a collection of pilots. Again, language matters. Pilots imply risk and failure. Use the term “demonstration” which signals confidence and reliable outcomes. Success requires engagement across clinical, operational, IT, security, compliance, and finance leadership. It also requires listening, especially to physicians and frontline staff who will ultimately determine whether these tools succeed or fail. It’s even better when you include patients and families in the design phase.
Be wary of vendor claims and industry buzz. Insist on evidence tied to your workflows, your data, and your outcomes. We have seen moments like this before with EHRs, interoperability, and analytics. What is different now is speed. The window between “emerging” and “expected” is shrinking. This moment is both exciting and dangerous. AI has the potential to improve patient and family experience, reduce clinician burden, enhance outcomes, and strengthen financial performance. It aligns directly with the Quintuple Aim.
Potential, however, is not value. Value comes from disciplined execution. Speed is secondary. The organizations that succeed will be the ones that move deliberately, align AI to strategy, govern it effectively, and measure what matters. Treat AI as the exciting and promising capability it can be, but do so with intention, not frenzy. That is a fundamentally different and far more promising approach.