When University of Rochester Medicine, a $6 billion academic health system serving a broad region of upstate New York, set out to deepen AI adoption, they didn’t look to a consulting firm or a centralized IT team. Instead, they looked to people who know the work best: clinicians, administrators, and analysts on the front lines.
The result was a six-week “Build Your Own AI Assistant” challenge: a structured, system-wide competition in which any University of Rochester Medicine employee could learn to design, build, and submit a custom AI assistant using Qualified Health's Assistant Builder platform. The numbers tell the story: more than 500 participants registered, 261 staff members built something, and the program yielded over 500 unique AI assistants.
500+ Registrants | 261 Builders | 500+ Assistants created
From Chat to Action
As a fully integrated academic health system, University of Rochester Medicine is the largest provider of healthcare throughout the Finger Lakes and Southern Tier regions of New York. It includes eight hospitals, hundreds of outpatient care facilities, urgent care centers, home care, long-term and continuing care facilities, and other programs. With more than 32,000 employees, it is the centerpiece of URochester’s medical research and health sciences education missions, including the School of Medicine and Dentistry, the School of Nursing, and Eastman Institute for Oral Health.
University of Rochester Medicine had already deployed Qualified Health’s conversational AI platform system-wide. But leadership wanted to surface the specific, high-value use cases that only frontline staff could identify and to turn those insights into deployable tools.
A build-a-thon model was the answer: University of Rochester Medicine invited all staff to become builders.
A Structured Path to Building
The program ran from mid-January through late March 2026. An open registration period drew more than 500 interested participants, including clinicians, researchers, program directors, project managers, and support staff. The formal build phase kicked off in early February with a weekly training series in which each session tackled a progressively more advanced topic: Assistant Builder basics, writing effective prompts, working with custom knowledge sources, handling complexity, testing and iteration, and, finally, assembling a polished submission.
Between sessions, participants had access to live office hours co-hosted jointly by Qualified Health and University of Rochester Medicine personnel, as well as a dedicated Microsoft Teams channel staffed by University of Rochester Medicine. That channel became a community hub in its own right, with 265 active users sharing their work, asking questions, and offering feedback to peers.
"There’s extraordinary expertise across our academic health system," notes Gregg Nicandri, MD, Chief Digital & Innovation Officer for University of Rochester Medicine. "The AI Assistant Challenge gave people the tools to put their expertise to work in new and innovative ways. What came back exceeded what any of us anticipated."
One participant, a procurement professional, shared an RFP Evaluator she built that assesses vendor responses, generates qualitative and quantitative analyses, and produces an executive summary with recommendations. A cardiologist and director of the Cardiology Data Analysis Unit contributed a tool capable of ingesting data on procedural volume and enabling conversational analysis to explain year-over-year trends by location, department, attending physician, and procedure type. Others built clinical tools: a new patient intake assistant that summarizes patient history for evaluation and documentation, a revenue capture tool to document MD and APP levels of service for RVU purposes, and a contracting assistant that reviews contract clauses and suggests more favorable terms.
Meet the Trailblazers
Concluding the six-week challenge, 46 participants submitted their custom-built Qualified Health Assistants for formal review. Representative of a broad range of the cohort’s work, eight of the submissions are now recognized as Trailblazers and can illustrate and inform what responsible AI governance looks like in practice. The job now is to develop the infrastructure that enables them for broader use.
The committee is working with each Trailblazer team to navigate the publishing process.
“These eight will shape how we support the next wave of builders,” notes Kathleen Fear, PhD, Senior Director of Digital Health and AI at University of Rochester Medicine.
More than a Technology Exercise
The participant response suggests the program’s impact ran deeper than the assistants themselves. In feedback collected during the challenge, staff described a shift in how they related to AI—moving from spectators to participants. The build-a-thon not only brought more team members to the AI platform but also significantly boosted their trust and engagement through hands-on tool development.
"I almost didn't sign up because I wasn't sure I could build an AI assistant," shares Karin Gaffney Christensen, University of Rochester Medicine Communications, editor of the twice-weekly Faculty Focus newsletter. "I'm incredibly proud of what I developed and the fact that it is exactly what I had envisioned. It saves me time and I believe could do the same for other University content editors using versions tailored to their specific needs."
Cross-functional collaboration was another unplanned dividend. The Teams channel facilitated connections between departments that might rarely interact otherwise, as a cardiologist’s data tool or a procurement team’s RFP evaluator drew questions and interest from colleagues across the system.
Governance as the Next Frontier
With submissions closed and initial winners announced on March 25, the challenge enters a new phase: governance review. One of the goals of the build-a-thon was to promote the most valuable assistants to the platform for broader deployment and use across the entire health system. Assistants selected for this broader deployment will go through a structured review process before being made available system-wide. This approach reflects the health system’s commitment to responsible AI adoption and one that other large health systems may look to as a model.
"We are humbled by the rigorous testing that is required," says Alexandra Yamshchikov, MD, Director of Outpatient Parenteral Antibiotic Therapy (OPAT) and Associate Medical Director, Antimicrobial Stewardship Program, of the GERMI team. "These are clinical recommendations that can impact patients in a very direct way, so they need to be correct and the tool needs to be rolled out with thoughtfulness."
The build-a-thon model itself offers a replicable framework for health systems wrestling with a common tension: the gap between enterprise AI deployments and the specific, granular workflow needs that only frontline staff can articulate. By pairing a no-code building platform with structured training and community support, University of Rochester Medicine found a way to make that gap productive and turn latent expertise into deployable innovation at scale.
"What excites me most about the Trailblazer cohort isn't any single assistant, it's what working through them together will teach us," Kathleen encourages. "Every governance question we answer, every pathway we build, makes it easier for the next great idea to find its way to the people who need it."
Qualified Health, the health AI company behind the platform, has positioned the Assistant Builder as a tool for distributed innovation. The University of Rochester Medicine build-a-thon represents one of the most extensive tests of that model to date.
.png)
.png)
.png)
