A conversation about how one ER and teaching physician navigates AI across three very different contexts — and what it reveals about how the rest of us should be thinking about AI.

Featuring Dr. Jennifer Campoli, DO, Residency Program Director and Associate Professor of Emergency Medicine at Upstate Medical University, Syracuse, NY

When I sat down and asked Dr. Jennifer Campoli how AI was showing up in her work, I expected to hear about clinical tools, workflow automations, and the medical research augmentation applications. Jennifer is an emergency medicine physician and Residency Program Director at Upstate Medical University, with over 20 years of experience. She runs a medical training program and sees patients in a level-one trauma environment at Upstate University Hospital. She has seen AI enter her field from multiple directions all at once.

What I didn't expect was that the moment in our conversation I'd still be thinking about days later had nothing to do with what happens inside the hospital at all — it was about what happens when someone who knows that world inside and out uses AI to help someone find clear and solid ground in a challenging health situation.

But to get there, let me first paint the backdrop and context for the story.

AI at the clinical level — the decision support tool

Inside the hospital, Jennifer's physician team uses a platform called Open Evidence for clinical decision-making at the point of care. If you haven't heard of it, the scale may surprise you. Open Evidence is used daily by over 40% of physicians in the United States, is active across more than 10,000 hospitals and medical centers nationwide, and draws from over 300 peer-reviewed medical journals, the FDA, and the CDC. It was developed by Harvard and MIT researchers and is designed to function, in their own words, “like having a curbside consult with a team of expert physicians available in your pocket.”¹²

What makes Open Evidence distinct from general-purpose AI — and what makes it trusted in clinical environments — is precisely what Jennifer identifies as its most significant limitation: it operates on a closed dataset. That is intentional. The platform is built for safety and clinical rigor, pulling exclusively from vetted, peer-reviewed sources. There are no hallucinations from the open web, and no unverified data slipping into patient recommendations.

For a physician trained to synthesize broadly — to draw from emerging research, cross-institutional findings, and sources outside a curated library — that boundary can be genuinely frustrating. The information available within the system is excellent. What cannot be cross-referenced is invisible, and you may not even know what you are missing.

How do you balance safety and control with the breadth of knowledge that genuine expertise requires?

AI at the administrative layer — the workflow tool

Upstate Medical University runs as a full Microsoft shop — meaning the entire organizational infrastructure, from communication to documentation to scheduling, lives within the Microsoft 365 ecosystem. For day-to-day administrative tasks, staff and physicians use Microsoft Copilot, which fits naturally into the Microsoft ecosystem.

This is AI as infrastructure rather than AI as intelligence. Drafting, summarizing, scheduling — the operational friction that accumulates in any large institution. It is a completely different tool serving a completely different need than Open Evidence, and it lives under the same roof. Two distinct AI layers doing two distinct jobs, within a single organization.

AI at the personal level — the advocacy tool

This is the part of the conversation I have not stopped thinking about.

Outside the hospital, Jennifer uses Claude — Anthropic's AI — not for clinical decision-making, but for something arguably more powerful. A friend of hers was navigating a difficult medical situation and heading into an appointment with specialists feeling overwhelmed and underprepared. Jennifer used Claude to generate a concise, agenda-style outline her friend could rehearse with and bring into the room: the right questions to ask, the right terms to understand - a simple framework for having an informed conversation with their physician rather than a passive one.

"She used it so her friend could walk into that appointment as a participant — not a patient who just receives information."

Consider what that means. A physician — someone who lives and breathes the medical system every day, who knows exactly how those appointments work and what gets said and what doesn't — reached for a consumer AI tool to help a friend navigate that same system from the outside. Not to replace the doctor. Not to self-diagnose. To close the gap between someone who knows how to advocate for themselves in a medical setting and someone who doesn't.

That is AI as an equalizer. That is the use case that doesn't make it into the press releases or the conference keynotes. And it may be the most important one.

What this tells us about where AI actually lives

What struck me most was realizing that Jennifer wasn't making arbitrary choices between tools. Each one mapped to a completely different context, need, and level of extraordinarily high stakes. Open Evidence for clinical rigor. Copilot for operational efficiency. Claude for human connection. The sophistication wasn't in using AI — it was in knowing which AI, and why.

That is something we don't talk about enough. AI fluency isn't just about knowing how to use a tool. It's about understanding that the tool should follow the use case, not the other way around. Role, context, and purpose should drive the decision — not defaulting to whatever is most familiar or most hyped.

In a single conversation with one person, I saw three completely different expressions of AI in the real world. That complexity is not a problem to be solved. It is the reality we are all navigating — and the sooner we name it clearly, the better equipped we will be to navigate it with intention.

How do you choose what AI tools to use for your work and life?

This post is part of the Women of Nymbl AI series — a campaign exploring real perspectives on artificial intelligence from the women building it at Nymbl. Follow along on LinkedIn using #WomenOfNymblAI.

‍