The Derby Mill Series

Managing AI Platforms for Healthcare (ep 27)

22 snips
Apr 28, 2026
Niamh Gavin, applied AI scientist exploring multi-agent systems and 360-degree healthcare. Sendhil Mullainathan, MIT professor studying ML and human decision-making. Tomi Poutanen, CEO building an AI management platform for hospitals. They discuss why hospitals need a control plane for many AIs. They demo observability, safety and governance tools. They debate evaluation, orchestration at scale, and consumer-focused healthcare visions.
Ask episode
AI Snips
Chapters
Transcript
Episode notes
ADVICE

Start By Inventorying Every AI

  • First get inventory: discover and catalogue every AI running in a health system before anything else.
  • Many AI tools are deployed locally by departments or researchers, so replace Excel spreadsheets with a central observability registry to expose duplication and risk.
INSIGHT

GenAI Will Take Low-Risk Patient Tasks First

  • Patient-facing GenAI can handle many routine interactions, but hospitals keep clinicians in the loop to manage liability and trust.
  • Typical low-risk tasks (appointment times, refills, addresses) could be automated once observability and trust metrics exist.
INSIGHT

AI Is Different Because It Degrades And Scales Widely

  • AI in healthcare needs special governance because it's probabilistic and can degrade, sometimes catastrophically.
  • Unlike deterministic software, thousands of narrow AI solutions will be required across diseases, transitions, communications, and operations.
Get the Snipd Podcast app to discover more snips from this episode
Get the app