The Pulse Library

The Strategy Is Not the Model

Technical teams build for clinicians instead of with them, governance drifts into an academic exercise, and the real risk sits in the gap between clinical and operations that neither one owns. Notes from the data side of healthcare on what actually moves adoption.

The Committee Nobody Staffed

AI oversight committees get announced, not staffed. What a functioning review body costs, what it produces, and the two questions boards should ask.

The Tools You Did Not Approve

Clinicians are already using AI tools nobody approved. Why bans fail, and what the hospital formulary teaches about the inventory first response.

When Accuracy Is Not Enough

A sepsis model that looked strong on paper missed most real cases. Why validation metrics are not value, and what leaders should demand instead.

The Vendor Owns Your Newsfeed

Much of what executives read about healthcare AI exists because a vendor paid for it. How to audit the feed, and where an independent briefing fits.

Best Healthcare AI Newsletters for Executives: An Honest List

The Population the Model Cannot See

The proxy beneath population health software decides who gets care. Why the target and the data, not the model, are the real work.

When the Software Stops Asking

Agentic AI does not hand the work back, it acts. What stays human, and what a health system still owns, once the software stops asking.

When the Patient Becomes the Training Data

Patient privacy is the AI exposure surface health systems are not watching. What leaders owe the person behind every record once AI tools enter care.