Most health system leaders can tell you
which clinical AI tools are deployed
in their institution.
Very few can tell you who owns the decision
when one of them is wrong.
The gap between those two answers
is not a governance failure.
It is a design failure.
And it starts before the contract is signed.
What a named Decision Owner actually is.
Not a committee.
Not a job title.
Not the vendor.
One person. Per deployment.
Someone who can be called when the algorithm
fires at 2am and the outcome is unclear.
That person needs two things to function
in that role.
An understanding of the data generating
the output. And the clinical guardrails
meant to contain it.
Without both, they own a title.
Not a decision.
Four substitutes appear in most
governance documents.
None of them hold.
The governance committee.
A committee deliberates.
A Decision Owner acts.
When a clinical AI output influences
a patient outcome in real time,
there is no time for a meeting.
Committees govern in advance.
They do not own the moment.
The CMIO title.
Organizational authority is not
deployment-level accountability.
A CMIO overseeing fifty clinical AI
deployments without a named owner
per deployment owns the policy.
Not the decision.
The vendor contract.
Your vendor is accountable for model
performance within defined parameters.
The clinical decision that followed
the output lives with your institution.
Unnamed.
The AI policy document.
Policy governs intent.
It does not name a person.
When discovery begins, a policy document
cannot answer a question.
A person can.
My years in cardiothoracic surgery
taught me that ownership is never assumed.
It is assigned before anyone
walks into the room.
The teams that performed well
had one thing in common.
There was always a plan
for when things went wrong.
Clinical AI is now in the room.
In most institutions,
the ownership conversation
has not happened yet.
Three gaps open when it does not.
The Regulatory Gap.
No named owner means no answer
when a regulator asks who owned
the clinical AI decision.
The regulator does not close that gap.
The institution does,
or the gap closes the institution.
The Operational Gap.
When the AI recommendation conflicts
with the clinician's judgment,
someone has to resolve it.
No named owner means the conflict
defaults to the clinician alone.
The clinician carries the clinical decision
and the unresolved accountability question.
That is not a workflow.
That is a burden transferred without consent.
The Legal Gap.
Litigation follows the document trail.
If the chart records the clinician's decision
and the AI's contribution is invisible,
the legal question lands on the clinician.
If the AI's contribution is documented
but no owner is named,
the legal question searches
for the institution.
The gap does not protect anyone.
It only relocates the exposure.
Naming one person per deployment
does not close all three gaps alone.
But nothing closes without it.
Someone always owns it.
The question is whether you named them first.
For every clinical AI deployment
currently live in your institution,
the named Decision Owner should be
identifiable without a meeting,
without a committee,
without a search through
a governance document.
If producing that name requires
any of the three,
the gap is open.
Name them before it matters.
MedicoVigilance™ is published every two weeks by Mo Johnson, MD MBA, founder of GPe Research. Each issue teaches one piece of the clinical AI accountability discipline your institution needs before the next deployment decision.
Forward this to a colleague whose institution is running on output no one has been named to own.