There’s no “AI” in connection

By Rahul Gorijavolu, medical student, BSPH, Johns Hopkins Medicine

It arrived the way the many insidious illnesses do—gradually, then undeniably. Years of deferred maintenance, symptoms dismissed or managed with quick fixes. In our diseased healthcare system, we built something that functioned just well enough for everyone to glance at the monitors and say it was stable. Until it wasn’t. 

The history of present illness 

It started with a single purpose—the alleviation of suffering. Our profession wasn’t built on technology. It was built on touch. One person showing up for another. 

But the anatomy changed. Reimbursement models, regulations, liability frameworks—none malicious, all necessary connective tissue for a growing body. But cumulatively, the tissue calcified, and the system remodeled itself around efficiency rather than healing, throughput rather than presence. The original purpose was buried under layers of administrative sclerosis. 

Now the vital signs are critical. Nearly half of physicians report burnout. We built a system that can monitor everything about a human body yet misses the human being before it. 

The diagnosis: by design 

The diagnosis isn’t moral. Healthcare professionals enter this profession to heal, only to find themselves bound by the same tissue that was supposed to hold the system together. The disease is in the structure. 

But call it architectural, and you let the architecture off too easily. The system didn’t calcify by accident. It calcified because efficiency is measurable and presence is not. Because throughput is billable and trust is not. Because every layer of administration that buried the encounter also generated revenue for someone. The incentives built this. Calling it a design flaw obscures the fact that for many of the people and institutions who shaped the system, it’s working exactly as designed. 

We built a container for medicine that’s too small for what medicine actually is. We defined value by what could be measured without measuring what matters most. There are no RVUs for listening, no quality metric for the moment a patient feels understood, no checkbox in the EHR for healing a person’s body, mind, and spirit. When the most important thing a clinician can do doesn’t exist inside the system’s definition of value, the system is the one that’s sick. 

The operation 

I spend my time working on AI research and governance. I’m also a medical student, learning how to care for patients at the exact moment we’re being asked to fix the system that’s supposed to care for them. AI is the most powerful instrument we’ve developed  in a generation, and we’re standing over an open patient. The question is what kind of operation this will be. 

I worry we’ll do what we’ve always done: treat the symptoms. Use AI to make the existing system faster—more documentation per hour, more patients per slot, more optimization of a model already failing the people inside it. That isn’t healing. That’s accelerating the disease. 

And there’s a darker possibility. Every tool we’ve introduced into healthcare—billing codes, EHRs, quality metrics—entered as support and calcified into structure. AI will face the same pressure. The same incentives that buried the encounter under administration will try to absorb AI into that same architecture. The system doesn’t lack intelligence. It’s unable to value what it can’t bill for. A smarter system with the same values isn’t a better system. It’s a faster one with the same disease. 

The real operation is structural. If AI can absorb the computational weight—the documentation, pattern recognition, inbox triage—then for the first time in decades, we have space to ask a question the system’s own mass has made impossible: What is the clinician actually for? 

I think the answer has always been the same: the irreducible act of one human being showing up for another. Knowledge, protocols, and pharmacology matter enormously. But they serve that central act. Somewhere along the way, we reversed the hierarchy, making the computational work primary and the human connection secondary. AI gives us a chance to put it back. But only if we also restructure the incentives that reversed it in the first place. Technology alone changes nothing. The fight is over what we choose to value. 

The prognosis 

The prognosis depends on what we do with this moment. AI isn’t the cure. It’s a brief window to pull back the scar tissue and see the original structure underneath. Letting it become another layer of administration is what we’ve done before and is the path of least resistance. 

Or we can operate. Rebuild the container around what called most of us into medicine—not efficiency, not throughput, but the human connection at the center of it all. We’ve never built a system that protects it. This could be our chance. 

The treatment plan is ours to write. Here are a few things to consider: 

1. When AI enters your workflow, ask what it’s making room for—and refuse to let the answer be more volume. 

2. Notice the moments that never make it into the EHR—the pause before a hard conversation, the shift in a patient’s body when they feel heard. Protect those moments. 

3. Resist the assumption that faster is better. Defend the space AI creates before the system fills it for you. 

4. Talk about this with your colleagues. Not what AI can do, but what medicine should be. 

This piece expresses the views solely of the author. It does not necessarily represent the views of any organization, including Johns Hopkins Medicine.