How AI Is Saving Healthcare | Shiv Rao - This Week in AI Ep. 5

By This Week in AI

Categories: AI

Summary

US healthcare spends 2x more than universal systems while delivering unequal outcomes—AI agents can solve the supply-demand crisis by automating 15-20% of costs tied to inefficient GP visits that average just 10-15 minutes. Founders should build AI intake systems with human-in-the-loop validation to replace low-value initial consultations.

Key Takeaways

  1. Doctors need 30+ hours of administrative work daily—pre-charting, pulling disparate data, writing notes, placing orders, coding diagnoses, getting insurance pre-auth. This represents massive automation opportunity for AI agents to handle backend workflows.
  2. 15-20% of healthcare system costs come from initial GP visits that are absurdly expensive for 10-15 minute encounters. AI-first triage with non-doctor human-in-the-loop can deflect these visits and reduce costs significantly.
  3. Two-tier patient outcome problem: top-tier hospitals have world-class technology, but rural patients drive 3-5 hours to reach care while local systems shut down. AI agents can democratize access by delivering standardized care in underserved areas.
  4. ChatGPT pre-visit behavior is already changing patient interactions—20% of patients now arrive as 'challenging' informed by AI, making them 100% of the doctor's cognitive load. Build AI systems that prepare patients AND doctors for smarter encounters.
  5. Foundation models can't legally give medical advice yet, but regulation ('it's coming') signals market opportunity. Build around compliance-first architectures with clear human-in-the-loop decision boundaries to prepare for inevitable regulatory frameworks.

Topics

Transcript Excerpt

Large language models can't give medical advice. >> It's coming. Doctors need 30 hours a day to get all of their work done. All of those jobs are what we're going after. >> Two choices. Go to the lower third of a general practitioners and get advice or get it from the top three or four models. Which would you rather see them do? >> I would always do the models. There's more demand to see clinicians. There's more people who are on chat GBT figuring stuff out that makes them think they've got to s...