Most hospital AI arrives top-down, procured or built by technical teams and handed to clinicians. In early 2026 Yonsei University Health System tried the reverse, running a month-long AI challenge in which frontline staff proposed and built agents for the operational problems they know best, mentored by the system's Digital Health Office on feasibility and commercial potential. Fifty-five employees on nineteen teams reached the finals, and eleven teams took awards.
The winning entries map the unglamorous terrain where hospitals actually lose time. The grand prize went to a procurement team whose chatbot fields the more than 2,000 purchase requests the system handles each year. Other honoured projects included an AI that helps cardiovascular nurses prepare for arrhythmia procedures, a tool supporting emergency nurses' report writing and incident analysis, a system that screens staff enquiries against personnel regulations, and an agent that automates nursing shift scheduling at the rehabilitation hospital.
Judged on practicality, creativity and direct usability in hospitals and offices, the challenge was explicitly a test of whether doctors, nurses and administrators can build working AI tools without being specialists, and the health system says the answer was yes, with plans to support the strongest projects towards commercialisation. For observers of digital health, the experiment is a data point on how AI capability diffuses through large clinical organisations: not only through flagship diagnostic models, but through dozens of small agents built close to the work.