I have always been someone who loves technology and is genuinely excited by what it can do for medicine. The promise of better efficiency, safer care, and more time with patients is hard not to believe in. Over the years, though, I have watched that promise fall short far too often. Again and again, I have seen powerful technology fail not because it lacked sophistication, but because clinicians were not meaningfully involved early in its design. The result is often a well built product that solves the wrong problem, or solves the right problem in a way that does not fit clinical reality, leading to tools that quietly go unused.
In the emergency department, care is unpredictable and often begins before a patient is formally registered in the system. Patients arrive critically ill, sometimes without identification, and decisions are made in real time. I have seen this play out with AI scribe tools, some of which were initially unusable because they required selecting a registered patient before documentation could begin. In practice, that requirement made little sense in an emergent setting. When EMS brings in a sick patient, care starts immediately. Once the workflow was changed to allow documentation to begin without choosing a patient first, the tool suddenly became usable. That single adjustment turned impressive technology into something that could actually support care.
Language access tools offer another example of how well intentioned design can fall apart in emergency settings. Many translation solutions rely on call lines or structured intake processes that require multiple steps before communication can begin. Sometimes that means entering patient identifiers or navigating prompts while a patient is unstable or in distress. In other cases, interpreters are available only through speaker phone lines that are difficult to hear in a loud emergency department. Tablet based translators can help, but they bring their own challenges around device availability, setup time, and audio quality. When workflows are not designed for urgency, communication and care suffers.
These problems are not unique to the emergency department. They reflect a broader issue in healthcare technology design. A primary care clinic operates on scheduled visits and continuity. An intensive care unit revolves around ongoing monitoring and detailed documentation. The emergency department functions in a constant state of uncertainty, interruptions, and rapid decision making. When developers assume a single workflow can serve all of these environments, technology that works well in one setting can become burdensome or unsafe in another.
This is why clinician feedback must be foundational rather than an afterthought. Frontline clinicians understand how care actually unfolds and where flexibility is essential. Emergency physicians can explain why rigid steps fail during critical moments. Outpatient clinicians can highlight efficiency needs and continuity of care. Intensivists can speak to the importance of precision and longitudinal tracking. These perspectives will not always align, and that diversity of input is exactly what leads to better design.
We have seen what happens when clinicians are not involved early. More than a decade ago, healthcare underwent a massive transformation with the widespread adoption of electronic medical records. Many of those systems were built around rigid rules, billing requirements, and compliance needs rather than clinical workflows. Today, they are consistently cited as a major contributor to clinician burnout. We are now entering another pivotal moment as AI and other technologies begin to reshape healthcare. If clinicians are involved from the beginning, these tools can make practice better rather than harder. If not, we risk repeating the same mistakes and spending the next decade wondering how promising technology once again became part of the problem instead of the solution.