Day 4 of my challenge to figure out how AI is disrupting different industries
Day 4 of 30: Healthcare 🏥
Law, finance, accounting. Now the industry where mistakes cost lives.
This is How AI is faring
The win: AI scribes that listen to the visit and write the note. 62% of hospitals on Epic had adopted one by mid-2025, up from almost none two years earlier. Cleveland Clinic cut note time by 14 min per clinician per day. Intermountain cut it 27% per appointment.
The mess-up: UnitedHealth is being sued over an AI tool (nH Predict) that allegedly denied post-acute care doctors had approved. The suit claims a 90% error rate, based on denials overturned on appeal. UnitedHealth denies wrongdoing and says humans made the decisions. Cigna faces a similar suit over claims allegedly rejected in ~1.2 seconds each. Both are unproven.
Who's ahead? Big, urban, system-affiliated. Predictive AI: 96% of large hospitals vs 59% of small ones. 86% of system hospitals vs 37% of independents. Urban 81% vs rural 56%. And it's wide but shallow: 100% of 43 big systems piloted AI scribes, only 14% fully deployed.
Tools: Microsoft's Dragon Copilot, Abridge, Ambience, and now Epic's own scribe.
Same LLMs as coding, but the output is a medical record. UCLA's trial found scribes sometimes left out info or got pronouns wrong. In radiology, one study found even experienced radiologists got worse when the AI was wrong.
The twist: in one study, AI lifted resident accuracy by 15-19 points. Better doctors, or doctors who never learn to see it themselves?
Doctors won't be replaced. The paperwork is. The real fight is AI allegedly overruling doctors at insurers.
Day 5 tomorrow.
@jahirsheikh8 I mean instead of asking the individual u can just directly ask clause and get the correct answer.
Wild times ahead for software development