Writings
Writings on AI, data, and the organizational realities behind success and failure.
Cleared for Deployment?
Aviation requires explicit authorization before entering controlled airspace. Clinical AI has no equivalent moment, even when models move into hospitals that differ fundamentally from those they were trained on.
Software Has Become Cheap. Institutional Knowledge Has Not
Agentic coding has made software dramatically cheaper to build. As implementation costs collapse, institutional knowledge, not code, becomes the scarce resource that determines whether transformation succeeds.
The Feedback Loop Nobody Built
Healthcare AI governance focuses heavily on deployment and validation. Without someone responsible for learning from real-world performance, clinicians become the only feedback loop that reliably functions.
Why the Person Holding the Map Still Matters
Models are useful precisely because they simplify reality. The real challenge is not building perfect maps, but knowing where they stop being reliable and when human judgment must take over.
The Validation Trap
Validation exists to reduce uncertainty, not eliminate it. When organizations treat it as a substitute for judgment, they exchange one kind of risk for another: avoiding decisions until uncertainty disappears.