22 percent.
That is how many hospital leaders
can hand a regulator
a complete AI explanation.
Auditable.
In 30 days.
According to Black Book Research.
The rest cannot.
As a surgeon,
I know what that number means.
A decision no one can account for.
Most health systems
do not govern clinical AI.
They perform it.
What most leaders think clinical AI governance is:
↳ Data governance with a new name
↳ Long policies no one enforces
↳ Banning tools and another approval layer
That is the performance.
None of it covers the decision.
Clinical AI does something
data governance never had to hold.
It makes a decision.
On a patient.
In real time.
A decision needs an owner.
This one needs two.
What clinical AI governance actually requires:
A Governance Owner sits above the deployment.
↳ Charter it
↳ Commission the owner
↳ Cover the institution when the model is wrong
A Decision Owner sits at the bedside.
↳ Decide with the model in hand
↳ Document what it shaped
↳ Defend the call
The Handoff sits between them.
↳ Where the AI stops and the clinician starts
↳ Name it, or the decision falls through
The Accountability Gap™ (TAG™) does not live
in the model or the data.
It lives in the Handoff.
If the clinician carries it alone,
you bought a workflow.
If the Handoff is named, with an owner
and a trail, you built governance.
Every health system
has a deployment running right now.
Most cannot name who owns it
when the model is wrong.
You already know whether yours can.
Mo Johnson, MD MBA is a cardiothoracic surgeon and the founder of GPe Research. Field Notes are short dispatches from the clinical AI accountability frontier, published alongside the MedicoVigilance™ newsletter at medicovigilance.org.
Follow the work on LinkedIn: linkedin.com/in/mo-johnson