Breast screening AI in a 105,000-woman trial raised cancer detection from 5 to 6.4 per thousand screened. Reading workload fell 44%. That’s real. But it’s not a 44% staffing cut.
Cardiology AI found more reduced ejection fraction. Gastroenterology AI found more polyps. Ophthalmology AI got screening to 100% completion vs 22% after referral. More detection creates downstream demand faster than capacity can grow.
Ambient scribes: 2.5 million encounters at one health system. Impressive adoption. A randomized trial across 238 physicians found one product cut note time 9.5%. Another: 1.7%, not statistically significant. Deployment volume is not a clinical outcome.
The pattern across all specialties: AI works best when the task has a defined input, a bounded output, and a clear owner for follow-up. The gap between a strong benchmark and a better patient outcome is real and measurable.
Subscribe to www.onhealthcare.tech for free and paid articles, podcasts, and more. For a further deep dive on the topic from today’s video teaser, see the podcast and article link in the comments.
