UHN:the Chief AI Scientist move
What they're doing
UHN appointed Dr. Bo Wang as Canada's first Chief AI Scientist at a hospital. The UHN AI Hub anchors cross-disciplinary work across cancer, cardiovascular, and transplantation. At the Princess Margaret Cancer Centre, the Cancer Digital Intelligence (CDI) program is reshaping cancer care with AI. In cardiovascular imaging, a model trained on 18 million videos from 300,000 UHN studies achieves 20% lower error on key measurements and can interpret pediatric scans without retraining. In radiation therapy, researchers have developed AutoPlanning, which produces treatment plans in minutes that used to take hours.
UHN's collaboration with Mayo Clinic on the global Platform_Connect initiative shares de-identified data across continents. The Michener Institute at UHN runs the AI in Health Care Certificate Program, training the next layer of practitioners.
Why it matters
UHN solved the org-design question most hospitals are still ducking. By appointing a Chief AI Scientist, they made AI a named function at the executive level, not a project inside IT or research. Every downstream decision about scale, governance, and partnership now has a clear owner.
What's working
- An executive-level AI mandate (Wang) backed by an operating unit (UHN AI Hub)
- Specific, measurable clinical wins (20% lower imaging error, AutoPlanning minutes-vs-hours) that scale across cancer, cardiology, transplantation
- Education pipeline (Michener Certificate) that produces talent for UHN and the broader Canadian health system
- International partnerships (Mayo, Cedars-Sinai) that pull in data and validation without giving up data sovereignty
What's harder than it looks
- The Chief AI Scientist role only works if the rest of the executive team is fluent enough to use it. Otherwise it becomes another silo
- Hospital-scale AI requires hospital-scale data. UHN has it. Most Canadian health orgs don't, which limits how transferable this model is
- International data partnerships create privacy and sovereignty tradeoffs that need to be communicated carefully to patients and the public
Three takeaways for similar orgs
- 01Name AI at the executive level. The title matters less than the mandate and the budget
- 02Build the training pipeline alongside the technical work. UHN trains people into the field, which feeds its own capacity and the sector's
- 03Smaller orgs that can't afford a Chief AI Scientist can still adopt the structural move: name the person, give them air cover, expect outcomes