Practical AI in Healthcare S1, E25 - Reflections 3: What Happens When Principles Meet Reality
Feb 22, 2026
They debate why their framework failed to capture recent stories and where their views diverged. They explore how unglamorous AI wins like billing can succeed while reimbursement often blocks practical impact. Legal and privacy rules already apply to AI and create liability and de-identification challenges. They warn about AI-enabled paper mills threatening the scientific record and discuss editorial guardrails.
AI Snips
Chapters
Transcript
Episode notes
Mundane AI Delivers Fastest Healthcare ROI
- Mundane AI wins often deliver the fastest, reliable business value in healthcare.
- SmarterDX turned clinical notes into billing codes by iterating from SQL to fine-tuned models that beat foundation models by ~30%.
Regulatory Approval Is Not The Final Barrier
- Getting regulatory approval isn't the only hurdle; aligning reimbursement and incentives is a separate 'second valley of death'.
- Alvin Liu's diabetic retinopathy screening works clinically but risks unsustainability if payment models don't adapt.
Current Reimbursement Penalizes Efficiency
- Fee‑for‑service reimbursement penalizes AI-driven efficiency because it rewards clinician time, not outcomes.
- Replacing a 20‑minute clinician task with a 30‑second AI analysis directly reduces billed revenue under current codes.