An AI-generated clinical plan can contain accurate findings, reasonable recommendations, and no obvious hallucinations—and still produce a poor plan for the patient in front of you.
That raises a patient-safety question every clinician must be prepared to answer:
Even if an AI-generated plan contains good ideas, are they the right actions for this patient, in this order, right now?
Consider this illustrative case.
A patient reports worsening fatigue, dizziness, and poor exercise tolerance.
Their records show anemia, low ferritin, positive thyroid antibodies, poor sleep, and gastrointestinal symptoms. They also report that starting several supplements together previously made them feel substantially worse.
AI might suggest an autoimmune protocol, gut testing, nutritional supplements, exercise, and sleep interventions. Some of those ideas may eventually be appropriate. The response could appear comprehensive and contain no obvious factual error.
Yet it has not resolved the most
important decision:
What should happen first?
The clinician’s first move is to assess the worsening symptoms and anemia, determine whether prompt medical evaluation is needed, and clarify the cause before beginning a broad protocol.
The patient’s previous intolerance provides another reason to avoid activating several interventions simultaneously.
The broader recommendations can be reconsidered after that assessment.
Worsening dizziness, fainting, breathlessness, or another
concerning change should trigger earlier reassessment rather than waiting for a routine follow-up.
This is clinical priority.
A patient may need many things—but they cannot all be first.
Many AI education programs teach valuable skills, including how to evaluate tools, check outputs, protect patient information, and maintain human oversight.
At Functional Medicine University, we also make the order of carean explicit part of clinical AI training.
We teach clinicians to determine:
What may be
urgent
What remains uncertain
What the patient can tolerate
Which action has earned the right to lead
What should deliberately wait
What evidence would require the plan to change
Because an AI answer can be accurate and still lead to a poor clinical plan when the priorities, timing, or sequence are wrong.
I’m sharing a sample lesson from
our FMU Clinical AI Certification™, together with its companion audio. The lesson introduces a practical priority check built around six questions:
As you evaluate any clinical AI training, I encourage you to ask:
Does it teach me only how to generate and evaluate an AI answer—or does it
also teach me how to determine what should happen first for the patient in front of me?
AI may identify the possibilities.
The clinician must determine the priority.
Want to Continue Following FMU's AI Work?
Over the next several weeks, I will be sharing additional white papers, audio commentaries, practical clinical AI guidance, and updates on the development of FMU's new AI Certification.
If AI is an area you want to follow more closely, you can join our dedicated FMU AI Updates List below.
Joining that list simply lets me know you want to receive our AI-specific educational updates rather than relying on the general FMU emails to catch them.
There is no obligation.
It is simply a way for those who are especially interested in AI and the future of clinical practice to stay connected to what we are developing.
I believe this is going to be one of the most important conversations in healthcare over the next several years.
And one question may eventually matter more than all the others:
Am I trained well enough to remain responsible while using increasingly capable AI?
I hope you find the sample lesson useful.
To your growth and success,
Dr. Ron Grisanti
Founder, Functional Medicine University
Dr. Robert
Conti
FMU Student Advocate
2123 Old Spartanburg Road #348 Greer South Carolina 29650 USA
