On 27 October 2026, the Association of American Medical Colleges published AI Competencies Across the Learning Continuum. For clinicians using AI in study or practice, it creates a timely opportunity to replace a vague aim of becoming better with AI with a learning task you can actually examine.
AAMC presents the work as an educational framework for curricula, professional development and formative assessment. It is not an announcement of a new licensing examination or a requirement to purchase an AI course. The practical approach below is iatroX editorial guidance for using the framework in your own learning.
Start with the breadth of the framework
The six domains cover foundational understanding, AI-supported clinical work, communication and collaboration, professional ethics, learning and improvement, and the wider systems in which tools operate. The three tiers are Novice, Proficient and Expert. AAMC explicitly separates these tiers from job titles: a physician can have different capabilities across domains, and expertise is not determined by seniority alone.
Use that breadth to choose a gap rather than trying to master every domain in a single session. You might understand a tool’s basic function but struggle to explain its limitations to a patient. You might be comfortable reviewing an answer but uncertain how to respond when the tool behaves unexpectedly within your organisation.
For a first learning plan, write the activity, the uncertainty and the evidence you would like to review. Keep your current role visible, but avoid making it the answer to the question of what you need to learn. A consultant and a student may both need to investigate an unfamiliar tool, while bringing different experience to the exercise. The examples below suggest exercises at increasing levels of responsibility; they do not establish or score your tier.
Novice: make independent reasoning visible
Choose an original educational case or a question you have not yet attempted. Before asking AI for help, write your answer and a short explanation. Include the finding you consider decisive and one plausible alternative. This creates a record against which you can examine the assistance you subsequently receive.
Then use the tool for a defined purpose. For example, ask which assumption in your explanation needs further investigation. Compare the response with your original account. Did it reveal missing information, introduce a new possibility, repeat what you had already said or make an unsupported claim? Write the observation before deciding whether to change your answer.
Review any decisive clinical claim against an appropriate current reference. The source should address the question and jurisdiction involved. If the reference is unclear or the disagreement remains unresolved, make that uncertainty the next teaching discussion. A longer generated explanation need not be the next activity.
Finish by explaining the decision again without copying the response. Keep the original and revised accounts together. They show what changed during that session, while a later unfamiliar case can help you examine whether you can apply the distinction independently. Neither account, by itself, establishes clinical competence.
Proficient: inspect a useful-looking answer
For someone who already uses AI regularly, choose a response that initially seemed helpful. Mark the sentences that influenced your next step. Then separate factual statements, interpretations and suggestions for action. You may find that only one or two claims carry most of the practical weight.
Check those claims first. Does the cited source exist? Does it support the statement being made? Does the answer preserve qualifications that matter to the case? Record the exact point of agreement or disagreement so that you can discuss it with a colleague. A generic note that the answer was accurate gives little direction for future review.
Next, inspect what information the response assumed. In a fictional consultation, it might have treated a patient preference as established when the scenario never supplied it. Your learning task could be to identify which question would clarify that assumption. Keep this separate from deciding whether a factual recommendation is current.
Repeat the exercise on a different topic after you have resolved the first issue. You are examining your review process, rather than awarding the tool a universal reliability score from two examples. The AMA’s existing AI principles also emphasise transparency about a tool’s intended use and evidence. Those principles are background professional guidance, rather than the new October announcement.
Expert: teach a decision about AI use
For a clinician with teaching or service responsibilities, choose one scenario in which learners disagree about how AI should be used. A useful discussion might concern when to seek an independent source, how to describe assistance in submitted work, or what to do when a tool’s suggestion conflicts with the information available.
Ask each participant to explain their initial decision before presenting a preferred approach. Make the disagreement specific enough to investigate. If the question concerns institutional policy, obtain the applicable policy. If it concerns clinical reasoning, use the relevant clinical evidence and teaching supervision. Keep a local rule distinct from a general educational preference.
Agree one change to the teaching activity and decide what you will observe afterwards. For example, you might ask learners to include a short account of which claim they checked and why. Review whether that account identifies meaningful uncertainty, rather than merely adding a sentence that says AI was used responsibly.
This is a proposed faculty exercise, not an AAMC assessment instrument. It can help organise a discussion and a follow-up review. It does not certify participants at an Expert tier or demonstrate that a particular product meets every competency in the framework.
Make your next CPD record about an observable change
Choose a single exercise and complete the follow-up before expanding the plan. A useful record might include your initial reasoning, the feedback or source you examined, the explanation you revised and the next situation in which you intend to use it. Retain an unresolved question when the activity reveals one.
Review the record later and ask what evidence is still missing. You may need a new case, a conversation with a teacher or a local process clarification. The answer determines the next activity. A polished reflection is less useful when it obscures the fact that the underlying question remains open.
iatroX Socratic Tutor supports reasoning review after an attempted question. The CPD workflow connects supported question activity with your reviewed reflection and learning record. These tools can provide a setting for the exercises above; no AAMC endorsement, competency certification or automatic accredited CME credit is claimed.
Frequently asked questions
Are the three tiers equivalent to student, resident and consultant?
No. AAMC explicitly describes developmental capabilities rather than a fixed mapping to professional roles. Use the domain relevant to your learning need.
Does the publication introduce a new US licensing requirement?
It presents an educational framework, rather than a new licensing examination or universal course requirement. Follow the requirements applicable to your institution and professional role.
Can an iatroX session certify that I meet the framework?
No such certification is claimed. A session can support practice and reflection, while formal judgments require the assessment arrangements appropriate to the capability being evaluated.
Sources and review
Reviewed 29 October 2026. Primary sources: AAMC competency framework, published 27 October 2026; AMA AI principles. The three exercises are original editorial suggestions, not official competency assessments.
Choose one AI-related learning gap: build a focused iatroX CPD session around an attempted question, the claim you need to check and the explanation you want to improve.