Fabio Ynoe de Moraes, Radiation Oncologist and Associate Professor at Queen’s University, shared a post on LinkedIn:
“GenAI in Oncology: Why most pilots die—and how to be the 5% that scales – lessons from other industries.
Hot take: GenAI isn’t failing because of ‘accuracy.’ It’s failing because it never survives the clinic.
In cancer care, extra portals = instant death. The winners redesign workflows, not just models.
Where GenAI already works (today):
- Tumor boards → 60–90s longitudinal briefs with guideline snippets (clinician edits = learning).
- Synoptic notes (RT/chemo) → draft + checklist verification of dose/fractions/OARs before sign-off.
- Prior auth → payer-specific letters in minutes; track denial → approval lift.
- Trial matching → I/E criteria parsed first; highlight exclusions to save time.
The 90-day rule:
- One job to be done (e.g., −25% time/case).
- Integrate into EHR/PACS/inbox—no new portal.
- Guardrails: PHI off by default, provenance on, human-in-the-loop.
- Measure absorption, not logins (how many steps did we delete?).
- Scale only if the 30/60/90 gates are met.
If we design for the clinic—not the demo—oncology can be the sector that proves GenAI’s value this quarter, not ‘someday.’
Your turn: What single oncology workflow would you automate first—and what metric would you bet on?”
More posts featuring Fabio Ynoe de Moraes.