AI Competencies Across the Learning Continuum

About AI Competencies

The Case for AI Competencies During These Dynamic Times

AI is rapidly reshaping health care. AI-enabled technologies now support clinical documentation, diagnostic interpretation, clinical decision support, patient communication, workflow optimization, research, education, and population health. As these capabilities continue to evolve, physicians must be prepared to critically evaluate AI tools, understand their strengths and limitations, recognize potential risks, and apply them responsibly in clinical practice.

AI is developing at a pace that challenges traditional approaches to establishing educational expectations and professional standards, and even our understanding of what it means to learn, practice, and care for others. Despite widespread interest in AI, there is currently no national consensus regarding what physicians should know and be able to do at different stages of their professional development. As a result, educational experiences vary considerably across institutions, creating inconsistent preparation for practice.

Developing a national set of AI competencies addresses this gap by providing a shared framework for educators, learners, institutions, and health systems. These competencies are intended to help ensure that physicians:

• Understand foundational AI concepts and terminology
• Critically evaluate AI-enabled tools before using them in patient care
• Recognize ethical, legal, regulatory, and professional implications of AI
• Identify and mitigate bias and unintended consequences
• Collaborate effectively with patients, colleagues, data scientists, engineers, and other professionals
• Integrate AI into clinical practice in ways that promote patient safety, health equity, and trustworthy care

Caveats

We heard from many through surveys, focus groups, and less formal methods that there are important caveats to consider as we navigate this period of change. This section contains a summary of the most common and consequential themes that emerged throughout the development process.

There is considerable promise for the role of AI in both education and health care. Used thoughtfully, AI may extend human capability, improve access to knowledge, support clinical reasoning and decision-making, reduce burdensome work, enable new approaches to education and discovery, and, ultimately, contribute to better care. Yet promise must not obscure peril. AI can reproduce or amplify inequities, introduce new forms of error and overreliance, threaten privacy, alter relationships and professional roles, and create incentives that may not always align with the interests of patients, learners, or communities. Its environmental, economic, labor, and societal consequences also warrant serious consideration.

Importantly, preparing physicians for an AI-enabled future must not mean diminishing the qualities that make medicine profoundly human. If anything, this period heightens the importance of the arts and humanities and of cultivating curiosity, interpretation, empathy, ethical reasoning, imagination, humility, and the capacity to sit with ambiguity. Technology may change how physicians acquire information or accomplish particular tasks, but it does not diminish the importance of understanding another person’s story, recognizing suffering, exercising judgment, building trust, or accepting responsibility for one’s decisions. The measure of progress should not simply be what AI allows us to do, but whether its use helps us become better physicians and provide more humane, equitable, and trustworthy care.

We strove to develop competencies that are not specific to a type of technology, AI product, or software application. The competencies are intended to reflect capabilities that will remain relevant as technologies continue to evolve. However, it is impossible to predict how rapid this evolution will be. Technologies will change, evidence will accumulate, policies and regulatory frameworks will mature, and our understanding of AI’s effects (both intended and unintended) will deepen. Some capabilities that seem essential today may become less important while others we cannot yet anticipate will emerge. Accordingly, these competencies should invite continued inquiry, evaluation, dialogue, and revision rather than suggest that the questions surrounding AI have been answered.

Although building competence in AI and other technological advances will remain a journey rather than a destination, this work aims to provide a standard set of expectations and outcomes along the learning continuum. We are all learners. New demands and advances in health care will require all health professionals, regardless of their level of training, to practice humility and maintain a growth mindset as they acquire new competencies and refresh existing ones. The developmental tiers (Novice, Proficient, and Expert) describe progressive levels of competence in the safe, effective, and responsible use of AI rather than specific learner or professional roles. Progression reflects an increase in knowledge, judgment, and the ability to apply AI across increasingly complex clinical, educational, research, and systems contexts.

Although the tiers are not intended to correspond directly to medical student, resident, or faculty roles, progression is influenced by experience, opportunities for practice, and increasing professional responsibility. Individuals may demonstrate different levels of competence across domains depending on their backgrounds and practice contexts. The Expert level (which may not be aspired to by all) generally reflects the integration of AI competencies with substantial clinical experience, educational responsibility, systems awareness, and professional judgment, rather than technical expertise alone.

Competencies related to system-level changes, policy changes, advocacy, and role modeling may be aspirational for some, depending on their local environment, education, and training. As with previous sets of competencies, we will strive to reach a consensus on balancing what some subject matter experts may consider below expectations and what others may consider aspirational. We will continue to listen to and learn from our diverse communities as we develop the next iteration.

We also recognize that engagement with AI, particularly large language models, is not universally embraced. Individuals and communities have raised serious ethical, environmental, creative, educational, and societal concerns about these technologies. Some have chosen to limit personal use. Such positions should not be dismissed as resistance to innovation. They represent perspectives worthy of respect, inquiry, and dialogue and can help illuminate consequences that enthusiasm for technological advancement might otherwise overlook. Meaningful consensus does not require uniform enthusiasm.

Organization of the Competencies and Intended Uses

Organization

The AI competencies were developed through a national, iterative consensus process involving experts in medical education, clinical care, AI, ethics, informatics, data science, and related disciplines. Literature review, structured stakeholder feedback, public comments, and focus groups informed successive revisions to the competency framework.

Consistent with the New and Emerging Areas in Medicine Series, the competencies are organized into developmental tiers that reflect progression across physician development. However, unlike previous sets, these tiers do not reflect a specific role or part of the continuum of medical education. The three tiers are:

• Novice
• Proficient
• Expert

Within each domain, competencies build across the continuum, with each developmental stage incorporating competencies achieved in previous stages while adding greater independence, judgment, leadership, and systems-level responsibility.

Intended Uses

These cross-continuum competencies are intended to help educators, institutions, and learners integrate AI thoughtfully into medical education and clinical practice. They are designed to supplement existing competency frameworks by providing greater specificity regarding the knowledge, skills, and professional behaviors required for responsible AI use.

Potential applications include:

• Engaging educators, clinicians, learners, patients, and interdisciplinary experts in collaborative discussions about AI
• Conducting gap analyses of curricula and training programs
• Supporting individual professional development and lifelong learning
• Developing curricular objectives and educational experiences
• Supporting the development of formative assessment strategies and tools
• Advancing research and scholarship related to AI in medical education
• Guiding the strategic integration of AI into curricula and clinical learning environments

The competencies presented in this report are community-informed consensuses on abilities related to AI in learning and practice. They are intended to foster responsible innovation while promoting patient safety, ethical practice, transparency, equity, and trust. The competencies go beyond physicians’ use of AI; they inform how physicians should exercise the wisdom, judgment, and humanity necessary to help shape how AI is (and is not) used in medicine. The competencies should also be understood not as a final destination, but as a snapshot of our collective thinking at this moment in time. These competences therefore represent an educational framework, an invitation, and a historical marker.

Competency Domain I: AI Knowledge for Practice

Explain what artificial intelligence (AI) is, how it works, and how its evolving tools and applications can be critically and responsibly used in health care, education, and research.

Table 1

Tier 1: Novice	Tier 2: Proficient	Tier 3: Expert
1a. Describes foundational AI terminology, concepts, and principles relevant to health care, education, and research, including the role of human judgment in AI-supported decisions and processes.	1b. Applies foundational AI knowledge to explain and evaluate AI capabilities, limitations, appropriate uses, and risks.	1c. Integrates and teaches foundational AI knowledge to support the appropriate use, verification, appraisal, governance, evaluation, and human oversight of AI in health care, education, and research.
2a. Describes how AI systems draw on biomedical, clinical, educational, and research data to generate outputs relevant to medical knowledge and practice, and how biases in these data and systems can perpetuate inequities.	2b. Evaluates AI-generated information for alignment with established biomedical knowledge, clinical evidence, and accepted standards of practice, and for potential bias or inequitable implications.	2c. Teaches and guides others in critically appraising AI-generated medical knowledge including how outputs reflect data sources, evidence quality, uncertainty, bias, and evolving standards of care.
3a. Creates AI inputs such as prompts, instructions, and contextual information, and explains how input quality and context influence AI-generated outputs.	3b. Iteratively refines AI inputs and critically evaluates outputs for accuracy, relevance, completeness, and limitations using human discernment and appropriate resources.	3c. Role models and teaches others effective interactions with AI systems, including strategies for creating AI inputs, verifying outputs, and critically appraising results using current best practices.

 

Competency Domain II: AI-Assisted Patient Care and Clinical Reasoning

Use AI with clinical discernment to assist and augment clinical reasoning, patient education, and care delivery while ensuring safe, equitable, patient-centered care.

Table 2

Tier 1: Novice	Tier 2: Proficient	Tier 3: Expert
1a. Explains that physicians remain accountable for decisions and actions supported by AI.	1b. Provides oversight for AI-supported clinical decisions and actions based on diagnostic stakes and clinical risk.	1c. Calibrates the use and oversight of AI based on clinical risk, uncertainty, patient complexity, and the potential consequences of error.
2a. Uses AI, when appropriate, to supplement an independently generated differential diagnosis and support initial management conclusions.	2b. Integrates AI selectively across the diagnostic and management processes while maintaining accountability for clinician-led reasoning.	2c. Teaches learners to use AI to augment, not replace, diagnostic thinking and management actions while preserving clinician agency, critical thinking, diagnostic humility, and independent judgment.
3a. Describes that AI outputs may contain errors, omissions, and bias, and may miss psychosocial, relational, and contextual factors.	3b. Identifies AI errors, omissions, and implausible outputs using clinical evidence, patient context, and independent judgment.	3c. Guides learner reflection to mitigate the impact of implausible or erroneous AI outputs on independent clinical reasoning and shared decision-making.
4a. Identifies clinical situations in which AI may appropriately support diagnostic or management decision-making.	4b. Selects and applies AI tools based on the clinical question, patient context, available evidence, and anticipated value to decision-making.	4c. Guides the use of AI across diverse clinical scenarios, determining when AI adds value, when it does not, and when independent clinical reasoning or alternative approaches are preferable.
5a. Detects uncertainty in medical decision-making arising from potential conflict between AI outputs and patient contextual factors.	5b. Reconciles discrepancies between AI outputs, clinical evidence, and patient presentation using independent clinical judgment.	5c. Models and teaches approaches for adjudicating conflicting AI-supported information while prioritizing safety, context, and clinician-led decision-making.

 

Competency Domain III: Communicating About AI and Interprofessional Collaboration

Communicate about AI-enabled processes, outputs, and limitations with patients, caregivers, colleagues, learners, and interdisciplinary collaborators to support understanding, trust, and responsible use.

Table 3

Tier 1: Novice	Tier 2: Proficient	Tier 3: Expert
1a. Describes which AI tools are used and their purposes and limitations in patient care, education, and research using language appropriate for the audience.	1b. Communicates how AI is being used in clinical decision-making, education, or research activities.	1c. Critically evaluates and guides best practices for communicating about AI use, including its benefits and limitations in clinical care, education, and research.
2a. Applies principles of responsible, safe, and effective AI-generated health information in patient care, including risks of misinformation and misinterpretation.	2b. Supports team members including patients and caregivers in the responsible, safe, and effective use of AI-generated health information, including information generated by consumer-facing AI tools and risks of AI-amplified misinformation.	2c. Leads organizational efforts to promote responsible, safe, and effective AI-generated health information, including strategies to educate and mitigate AI-amplified misinformation.
3a. Identifies when informed consent or disclosure is required for AI use and when patients or caregivers need additional education to understand the role of AI in their care.	3b. Discloses use of AI tools for patient care, education, and research and assesses understanding to obtain informed consent when applicable.	3c. Develops guidelines and regulations for best practices for disclosure, informed consent, and patient-centered communication about AI use and misuse.
4a. Differentiates the roles of health care professionals and nonclinical experts in AI-enabled health care.	4b. Works effectively with interprofessional and interdisciplinary teams to evaluate, implement, and optimize AI-enabled health care solutions.	4c. Leads collaboration across clinical, technical, operational, and ethical domains to advance the responsible use of AI and improve patient and system outcomes.
5a. Recognizes that AI-assisted documentation carries the same legal, ethical, and professional accountability as manually authored documentation.	5b. Edits and refines AI-generated documentation to correct errors, unsupported information, or misleading language; prevent inaccurate or fraudulent documentation or billing; and accurately represent clinical reasoning and medical decision-making.	5c. Supports documentation standards and clinical cultures that prioritize human review, accuracy, and reasoning over efficiency metrics in AI-assisted documentation.

 

Competency Domain IV: Professionalism and Ethics

Demonstrate ethical, legal, social, and professional awareness in the development, deployment, and use of AI, applying humanism and the principles of autonomy, beneficence, nonmaleficence and justice while preserving the professional identity, responsibilities, and trustworthiness of physicians.

Table 4

Tier 1: Novice	Tier 2: Proficient	Tier 3: Expert
1a. Explains how bias may be introduced into AI models through data selection, design, and implementation and how this may impact patient care.	1b. Evaluates AI-generated outputs for potential inaccuracies, bias, inequitable access, and impacts on disadvantaged populations.	1c. Promotes interprofessional collaboration to support equitable, harm-reducing AI development, implementation, and evaluation.
2a. Articulates how AI use in health care may contribute to environmental impacts, including energy consumption, electronic waste, and resource utilization.	2b. Applies principles of AI stewardship by using AI tools responsibly and avoiding unnecessary or excessive use in academic, clinical, or research tasks.	2c. Mentors learners and colleagues in responsible, resource-conscious AI practices and participates in efforts to monitor, report, and mitigate environmental harms associated with AI use.
3a. Describes ethical and professional considerations related to AI use in research and scholarly work, including core principles of transparency, authorship, accountability, protection, and integrity.	3b. Uses AI tools for research and scholarship in accordance with core ethical research principles, institutional policies, and journal guidelines.	3c. Advocates for development and use of AI tools in academic work that maintain professional boundaries and ethical standards.
4a. Explains that AI-enabled practice changes the physician’s work but does not replace core professional responsibilities, including humanism, accountability, integrity, and patient trust.	4b. Demonstrates professional identity in AI-enabled environments by exercising discernment, restraint, accountability, and compassion when using AI in care, education, or research.	4c. Models and teaches how physicians preserve professional identity, humanism, trustworthiness, and ethical responsibility while leading AI-enabled care, education, and scholarship.

 

Competency Domain V: Practice-Based AI Learning and Improvement

Uses AI to support lifelong learning, reflective practice, critical thinking, and improvement while maintaining expertise and humility.

Table 5

Tier 1: Novice	Tier 2: Proficient	Tier 3: Expert
1a. Delineates how AI can support individualized learning, self-assessment, and identification of personal knowledge and skills gaps.	1b. Uses AI tools to support learning goals within clinical skills, scholarly work, and professional practice.	1c. Evaluates and teaches how AI tools can support learner development, professional growth, and recognition of individual limits in knowledge, skills, or judgment.
2a. Differentiates how AI can support, or weaken, critical thinking, reflection, and foundational skills development.	2b. Uses AI to support reflective practice, critical thinking, and continuous learning to promote upskilling while monitoring to avoid deskilling or over-reliance.	2c. Engages in reflective practices, with self and others, on effective use of AI for continuous learning and improvement that mitigates deskilling and over-reliance.
3a. Explains the need to question one’s understanding, identify uncertainty, and seek credible sources when AI-supported information is incomplete, unclear or unfamiliar.	3b. Uses professional judgment and humility to question understanding, identify uncertainty, and seek credible sources when AI-supported information is incomplete, unclear, or unfamiliar.	3c. Models and teaches approaches to identify uncertainty, recognize knowledge gaps, and seek credible sources to improve AI-enabled practice.
4a. Identifies risks of over-reliance on AI for humanistic practice and potential never-skilling of essential clinical skills.	4b. Incorporates AI tools in ways that preserve and strengthen skills, confidence, and humanistic practice.	4c. Role models and teaches strategies to prevent automation complacency and preserve humanistic care.

Competency Domain VI: Systems-Based AI Practice

Evaluates, monitors, and improves AI integration across health systems to enhance quality, safety, access, and equity through physician engagement, co-development, critical appraisal, and safeguards.

Table 6

Tier 1: Novice	Tier 2: Proficient	Tier 3: Expert
1a. Explains governance, regulatory, legal, liability, clinical accountability, and physician autonomy considerations related to AI-enabled tools in patient care.	1b. Evaluates AI-enabled tools for governance, regulatory compliance, validation, liability, accountability, and physician autonomy prior to use.	1c. Collaborates with institutional governance and multidisciplinary partners to co-develop, critically appraise, and refine AI-related policies or practice concerns that affect patients, clinicians, colleagues, learners, or the environment.
2a. Describes privacy and security risks associated with AI tool use, both broadly and specific to institutionally nonsanctioned AI use.	2b.  Adheres to institutional policies for sanctioned AI tools and applies appropriate safeguards for protecting health and other sensitive information.	2c. Advises colleagues and learners on appropriate policies for AI tool use including addressing misuse or abuse.
3a. Delineates institutional processes for reporting AI-related safety events, adverse outcomes, and unexpected behaviors, or autonomy-limiting uses of AI.	3b. Reports AI-related safety events and concerns through institutional reporting systems, including unexpected behavior or model drift.	3c. Contributes to institutional learning, innovations, and quality improvement from AI-related safety events and to local efforts that improve safe AI use and preserve appropriate physician agency at a systems level.
4a. Describes how system-level AI implementation choices can contribute to inequitable outcomes, including impacts from algorithmic performance variation and from disparities in access, infrastructure, usability, or burden.	4b. Identifies and reports inequitable AI performance, access, usability, burden, or outcomes across patient populations within one’s practice setting.	4c. Leads or supports equity focused assessment of AI implementation across patient populations and advocates for system-level changes to address inequitable performance, access, burden, or outcomes.
5a. Explains how AI tools may alter clinical workflows, team dynamics, documentation practices, workload, accountability, physician autonomy, and patient experience.	5b. Participates in evaluating and refining AI-enabled workflows,  recognizing both intended and unintended effects of AI-integration on their care delivery, team roles, workload, documentation, patient experience, and environment.	5c. Contributes to systems-level, AI-enabled workflow design or refinement and monitoring of AI integration, accounting for sociotechnical, operational, educational, equity, safety, quality, and environmental implications.

To download the complete set of competency tables, a reference summary is available.

AI Curricula Profiles

This section includes brief profiles of AI curricula that demonstrate many innovative approaches to teaching and learning about AI.

Acknowledgments

We deeply appreciate all who contributed their time, expertise, and insights to advance education for the responsible use of artificial intelligence across the medical education continuum. The development of this competency framework was a highly collaborative, iterative effort, reflecting the collective expertise, perspectives, and contributions of a broad community. The AAMC gratefully acknowledges the project team, who led and coordinated this work; the expert members of the AI Advisory Committee, who provided guidance on the competencies; and the AI Working Group, who played a central role in developing and refining successive drafts. Hundreds of students, residents, faculty, affiliates, patient advocates, and others contributed feedback throughout the process, informing deliberations and helping build consensus. The resulting framework represents a shared body of work shaped by these diverse perspectives.

AI Advisory Committee, 2025-2026

J. Sybil Biermann, MD
Professor of Orthopedic Surgery
Associate Dean of Graduate Medical Education
University of Michigan Medical School

Christy Boscardin, PhD
Director of Artificial Intelligence and Student Assessment
University of California, San Francisco, School of Medicine

Morgan Cheatham, MD
Partner and Head of Healthcare and Life Sciences, Breyer Capital
Pediatrics Resident, Boston Children’s Hospital

Mark S. Cohen, MD
Dean and Professor of Surgery, Oncology, and Biomedical & Translational Sciences
Carle Illinois College of Medicine
Founder Professor of Bioengineering, Grainger College of Engineering
University of Illinois Urbana-Champaign

Laura Edgar, EdD, CAE
Senior Vice President of Competencies, Milestones, and Faculty Development
Accreditation Council for Graduate Medical Education

Regina K. Fleming, DO
Assistant Dean of Osteopathic Principles and Practice/Osteopathic Manipulative Medicine
Baptist University College of Osteopathic Medicine

Elissa Hall, EdD, MA
Director of Advanced Digital Education
Mayo Clinic

Cornelius A. James, MD, FACP, FNAP
Associate Professor of Internal Medicine, Pediatrics, and Learning Health Sciences
University of Michigan Medical School

May Lin, DO
Assistant Dean of GME, Technology, and Program Development
Academic Advisor, One Medical Group
Technology Advisor, American Association of Colleges of Osteopathic Medicine
Touro University, California, College of Osteopathic Medicine

Kimberly Lomis, MD
President
Josiah Macy Jr. Foundation

So-Young Oh, MS, MA
Director of Division of Digital Learning
NYU Langone Health

Martin Pusic, MD, PhD
Director of Research and Education Foundation
American Board of Medical Specialties
Associate Professor of Pediatrics and Emergency Medicine, Harvard Medical School

Adam Rodman, MD, MPH, FACP
Director of AI Programs
Beth Israel Deaconess Medical Center

Steve Singer, PhD
Vice President of Education and Outreach
Accreditation Council for Continuing Medical Education

Mark Speicher, PhD, MHA
Senior Vice President of Research, Learning, and Innovation
American Association of Colleges of Osteopathic Medicine

Yating Teng, PhD, MS
Director of Educational Outcomes
American Association of Colleges of Osteopathic Medicine

Victoria Yaneva, PhD
Director, Data Science and AI
National Board of Medical Examiners

AI Working Group, 2025-2026

Jennifer Allen, EdD, MEd
Instructor of Medical Education
Mayo Clinic College of Medicine and Science

Charles Binkley, MD, FACS
Director of AI Ethics and Quality
Associate Professor of Surgery
Hackensack Meridian School of Medicine

Rachael Callcut, MD, MSPH, FACS 
Associate Dean of Data Science and Innovation
Professor of Surgery
University of California, Davis, School of Medicine

Candace Collins, MD
Pediatric Critical Care Medicine Fellow
Johns Hopkins Children’s Center

Susannah Cornes, MD
Professor of Clinical Neurology
University of California, San Francisco, School of Medicine

John Cunningham, MD, FACP
Clinical Associate Professor of Internal Medicine
Interim Assistant Dean for Preclinical Curriculum (UME)
University of Texas Health Science Center at San Antonio

Alex Dummett, MD, FACP, CPPS, CPHFH
Chief Executive Officer
Socratic AI

Evita Henderson-Jackson, MD
Professor of Pathology
University of South Florida Morsani College of Medicine
Senior Member, Anatomic Pathology
Moffitt Cancer Center

R. Logan Jones, MD, FACP
Associate Professor of Medicine
Chair, Entrustment Group
Oregon Health & Science University School of Medicine

Daniel Lai, MD
Internal Medicine Chief Resident
Penn Medicine

Anna Maw, MD, MS
Associate Professor of Medicine
University of Colorado School of Medicine at the Anschutz Medical Campus

Scott Moerdler, MD
Associate Professor of Pediatrics
Fellowship Program Director, Pediatric Hematology Oncology
Rutgers Cancer Institute
Rutgers Robert Wood Johnson Medical School

Sara Nelson, MD, MHPE
Associate Professor of Emergency Medicine
Director, MaineHealth Maine Medical Center Medical Education Fellowship
Tufts University School of Medicine

Andrew Parsons, MD, MPH, FACP
Associate Dean for Clinical Competency
Division Director, Research and Academic Advancement
Associate Professor, Division of Hospital Medicine
University of Virginia School of Medicine

Cole Pieroni
Medical Student
Virginia Tech Carilion School of Medicine

Stacey Pylman, PhD
Associate Professor
Director of Continuing Medical Education and Educator Development
Michigan State University College of Human Medicine

Neria Sebastien, EdD
Director of Faculty Programs and Services
Center for Teaching and Learning
Columbia University Irving Medical Center

Andrew Tai, MD, PhD
Associate Professor of Internal Medicine and Microbiology and Immunology
Assistant Dean for Early Medical Education
University of Michigan Medical School

Shivam Vedak, MD, MBA
Associate Director for AI in Medical Education
Clinical Assistant Professor of Medicine
Stanford University School of Medicine

Julie Youm, PhD
Associate Dean of Education Compliance and Quality
Director of Educational Technology
University of California, Irvine, School of Medicine

AAMC Operational Team

Virginia Bush, PMP
Senior Project Management Specialist

Katrice Cain, MA
Director, Patient and Community Partnership Development
Patient, Family, and Community Engagement

Diane Cassidy, MLI
Director, Reference Center and Archives

Anne Farmakidis, MPS
Senior Director, Digital Medical Education

Toni Gallo, MA
Director, Medical Education Digital Strategy and Engagement

Lisa Howley, PhD, MEd
Senior Director, Transforming Medical Education
Co-PI and Project Lead

Mallory Lee, MEd
Business Operations Specialist

Dallas Peoples, PhD
Program Specialist, Strategic Initiatives and Partnerships

Meaghan Quinn, MHSA
Manager, Ambulatory Access and Population Health
Advancing Clinical and Operational Leadership

Alexis Rossi, PhD
Director, MedBiquitous^®

Beatrice Schmider, MA
Manager, Executive Operations

Kamilah Weems, MS
Director, Strategic Initiatives and Partnerships
Project Co-lead

Alison Whelan, MD
Chief Academic Officer, Academic Affairs
Co-PI and Executive Sponsor

Appendix

Appendix A. Development Process and Survey Questions

The AI competencies were developed through an iterative, evidence-informed review process that utilized multiple forms of stakeholder feedback. While no single standard approach exists to developing competencies, some preferred steps include those listed below. As in other works in this series, the AI competencies were developed following these general steps:

1. Define the scope of the competencies.
2. Engage diverse stakeholders and collect data to understand the nature of the construct. How do high performers perform? How do those who achieve positive results behave in practice? Gather data through literature reviews, focus groups, and electronic surveys of subject matter experts.
3. Draft the competencies.
4. Have reactor panels composed of subject matter experts, patients, frontline clinicians, and educators review multiple drafts of competencies.
5. Importantly, review and update competencies periodically to keep pace with changes in clinical and educational practices.

In September 2025, the AAMC convened the AI Competencies Working Group and charged them with designing, drafting, and refining a rigorous, national, consensus-based set of AI competencies for medical education. In close collaboration with the Advisory Committee and AAMC Operational Team, the Working Group was expected to use evidence-informed and transparent decision-making, participate in structured consensus activities, conduct iterative drafting and review, incorporate feedback from focus groups and multiple public calls for comment, and identify considerations related to implementation, assessment, feasibility, accessibility, potential harms, and unintended consequences. The deliverable would be a publishable competency set created by the AAMC in consultation with the Working Group and informed by community input, intended to support curricular gap analyses, individual professional development, the creation of learning objectives and assessment tools, research and scholarship, interdisciplinary dialogue, and the strategic integration of AI into formal curricula and clinical learning environments. This group of 20 met virtually from October through November 2025 to review an AAMC-conducted literature review. The Working Group drafted an initial set of six domains of competence based on their experience, expertise, and literature review. Working Group members were then divided into six smaller groups that met from December 2025 to February 2026 to define tiered competencies within their assigned domain of competence.

Once the initial set of competencies were drafted, several approaches to collecting stakeholder feedback were taken. A modified Delphi approach was conducted, which included two iterative rounds of stakeholder review. Separate electronic surveys were developed with standard items (see Survey Questions) and distributed to reactor panelists via invitational survey and the general public via anonymous open comment survey link. The first round of surveys distributed in March 2026 sought feedback on the initial draft domains and competencies, with 683 stakeholders in total responding, including 66 reactor panelists. The second round of surveys was distributed in May 2026 and evaluated a revised draft informed by feedback from the first round, with responses from 427 total stakeholders, including 52 reactor panelists.

Surveys

Respondents reviewed each competency domain and its associated subcompetency statements before answering the following standard questions:

1. Importance: Does this domain represent the important abilities needed for this domain?
2. Tier: Does this domain reflect observable abilities at the appropriate level for:
   • Entering residency (recent medical school graduate)
   • Entering practice (recent residency graduate)
   • Experienced faculty physician (3-5 years post-residency)?
3. Comprehensiveness: Does this domain reflect all specific abilities that are relevant to this domain?
4. Granularity: Does the language reflect an optimal level of granularity (neither too specific nor too broad)?
5. Relevance: Does this domain represent abilities that are relevant for all physicians regardless of specialty and/or practice setting?
6. Please provide general feedback on this domain or elaborate on any of your ratings above.

After reviewing the complete set of draft AI competencies across all domains, respondents were asked to answer the following two questions:

1. What, if anything, is missing?
2. How might the set of competencies be used to advance this emerging area in medical education?

Focus Groups

We also collected feedback through six focus groups at the 2026 AAMC spring meetings, including the Northeast Group on Educational Affairs Annual Conference, Together We Lead Conference, Central Group on Educational Affairs Annual Conference, International Advisory Committee for Artificial Intelligence, and Information Technology in Academic Medicine Conference. Focus groups were semistructured and included the following questions:

1. Importance: Does this set of draft competencies represent the knowledge and abilities needed for high-quality, safe, responsible, and ethical use of AI in health care?
2. Tier: Do these competencies reflect observable abilities at the appropriate level for:
   • Entering resident (recent medical school graduate)
   • Entering practice (recent residency graduate)
   • Experienced faculty physician (3-5 years post-residency)?
3. Comprehensive: Does this set of competencies capture all relevant abilities?
4. Granularity: Does the language reflect an appropriate level of specificity and detail?
5. Relevance: Are these competencies relevant for all physicians, regardless of specialty or practice setting?
6. Educational Usefulness: How could these competencies be improved to better advance the goals of medical education related to AI?
7. What, if anything, is missing from the competency set?
8. Is there anything else you would like to share that we did not ask about?
9. Are there specific competencies you believe should be revised or prioritized?

Note: All questions in bold required a response; all others were optional.

The Revisioning Process and Role of AI

Successive drafts were reviewed through early-reactor surveys, focus groups, advisory-group input, and broader public comment, with quantitative ratings of importance, comprehensiveness, granularity, relevance, and developmental level complementing qualitative feedback. Feedback from all sources was synthesized into cross-cutting themes and domain-specific changes, including identifying missing concepts; reducing duplication; clarifying the “natural home” of competencies across domains; strengthening attention to human judgment, accountability, equity, governance, implementation, and faculty development; and improving the ability to observe and assess competency statements.

The developmental framework was also reconsidered, moving toward more flexible novice-to-advanced expectations rather than rigidly tying competence to undergraduate medical education, graduate medical education, or faculty roles. The later review (Draft 2) shifted from identifying major content gaps to refining language, developmental progression, domain boundaries, and implementation considerations.

AI supported this human-led process, particularly as a tool for iterative synthesis, critique, and refinement. It was used to examine draft competencies for gaps and redundancies, compare related statements across domains, propose clearer or more durable wording, generate and refine developmental performance descriptors, and stress-test whether competencies would remain relevant as specific technologies change. The Working Group remained responsible for judgment and interpretation while final edits and decisions were determined by the AAMC project leads. In this sense, the development process itself modeled an important principle embedded in the competencies: AI can augment complex professional work, but human expertise, deliberation, and accountability remain central.

Appendix B. Glossary

artificial intelligence (AI): A machine-based system that can — for a given set of human-defined objectives — make predictions, recommendations, or decisions influencing real or virtual environments. Artificial intelligence systems use machine- and human-based inputs to perceive real and virtual environments, abstract such perceptions into models through analysis in an automated manner, and use model inference to formulate options for information or action.²²

AI foundational knowledge: Includes what artificial intelligence is, how it works, and how its evolving tools and applications can be critically and responsibly used in health care, education, and research.

AI-amplified misinformation: False, inaccurate, or misleading information, of which the creation, personalization, dissemination, reach, speed, apparent credibility, or impact is increased through the use of AI.¹⁹

AI-assisted (assistive AI): AI-enabled products designed to assist human decision-making. The AI only provides suggestions, information, or data that helps users make more informed decisions.²³

AI-enabled: A computer system that uses AI techniques to perform tasks that typically require advanced deductive or inductive tasks based on collected data of all types. These tasks include, but are not limited to, image and video analysis, predictive analytics, autonomous control and decision-making, and language and speech processing. AI-enabled systems have the potential to augment and enhance human capabilities.¹⁴

AI-generated (generative AI): AI models; in particular, those that use machine learning (ML) that are trained on large volumes of data and are able to generate new content.⁴

AI input: Data, information, instructions, or other content provided to or accessed by an AI system to generate an output.¹⁶

AI model: A core component of an AI system used to make inferences from inputs to produce outputs. ¹⁶

AI output: (Generated) content created by the model based on patterns in training data, the prompt, and tool settings. (Retrieved) content pulled from identified sources, such as websites, documents, databases, articles, or connected files.⁷

AI stewardship: Ethics embodied in the careful and responsible management of the design and use of AI.¹⁵

AI-supported: An approach in which artificial intelligence is used as a supporting or collaborative tool rather than a replacement of the human learner or professional. In education, AI-supported learning positions the learner as a collaborator with AI, using technology to assist learning, reasoning, feedback, or other activities while maintaining human participation and agency.¹⁷

AI system: A machine-based system that, for explicit or implicit objectives, uses the inputs it receives to determine how to generate outputs such as predictions, content, recommendations, or decisions that can influence physical or virtual environments. ¹⁶

autonomous AI: AI-enabled products that have the ability to perform tasks, operate independently, and make decisions without human intervention. The level of autonomy can vary based on the product.²²

competency-based medical education: “An outcomes-based approach to the design, implementation, assessment, and evaluation of medical education programs, using an organizing framework of competencies.”³

consensus: A form of collective agreement reached by members of a group after considering the relevant information, arguments, and perspectives.¹²

deskilling: The loss, deterioration, or reduced use of previously acquired skills and expertise, often because technology or changes in work processes take over tasks that people formerly performed themselves.¹¹

ethical standards: The guiding principles and regulatory frameworks that ensure AI tools are integrated into teaching, learning, and patient care with honesty, safety, and fairness.¹

equitable and harm-reducing AI: Creating and using technology that is fair for everyone and actively prevents negative or dangerous outcomes for students and patients.

explicit bias: “The traditional conceptualization of bias. With explicit bias, individuals are aware of their prejudices and attitudes toward certain groups. Positive or negative preferences for a particular group are conscious.”²⁴

health equity: The principle underlying a commitment to reduce — and, ultimately, eliminate — disparities in health and in its determinants, including social determinants.²⁶

implicit bias: “[Also known as unconscious bias,] refers to attitudes or stereotypes that are outside our awareness but nonetheless affect our understanding, our interactions, and our decisions.”²

misinformation: False, misleading, or inaccurate information that is shared without verified intent to deceive.

never-skilling: In the context of AI and medical education, never-skilling refers to the risk that learners fail to develop important foundational skills in the first place because they rely on AI during the period when those skills would normally be acquired.⁸

professional boundaries: “The evolving ethical, relational, and operational limits that define where algorithmic automation ends and human clinical judgment, accountability, and empathy begin.”⁹

professional identity formation: “The ongoing, deliberate process through which individuals integrate the values, norms, and behaviors of their profession into who they are.”¹⁰

prompt: An instruction, question, command, or other input provided to a generative AI system to guide its response or behavior.²⁵

reflective practice: “The ability to reflect on one's actions so as to engage in a process of continuous learning.”²⁰

role model: “One who serves as an example for others by demonstrating the behavior associated with a particular position or profession.”¹³

upskilling: The process of developing, improving, or updating an individual's existing knowledge, skills, and competencies, so that the person can perform current or evolving tasks more effectively.⁵

References

1. AAMC. Ensure ethical and transparent use.  Published July 31, 2025. Accessed Aug. 24, 2026. https://www.aamc.org/about-us/mission-areas/medical-education/principles-ai-use/ensure-ethical-and-…
2. AAMC and The Kirwan Institute for the Study of Race and Ethnicity at The Ohio State University. Proceedings of the Diversity and Inclusion Innovation Forum: Unconscious Bias in Academic Medicine. AAMC; 2017. https://store.aamc.org/downloadable/download/sample/sample_id/168/
3. Frank JR, Snell LS, ten Cate O, et. al. Competency-based medical education: theory to practice. Med Teach. 2010;32(8):638-645.
4. Harris L. Generative artificial intelligence: overview, issues, and considerations for Congress. Congress.gov. Published April 2, 2025. https://www.congress.gov/crs-product/IF12426.
5. International Labour Organization. Glossary of skills and labor migration. Published Dec. 3, 2021. Accessed August 2026. https://www.ilo.org/resource/glossary-skills-and-labour-migration
6. Cunningham JM, Jin L, Kazi I, Skandhan A, Rodriguez R. Improving resident knowledge of artificial intelligence ethics and prompting for clinical use. Clin Teach. 2026;23(4). doi:https://doi.org/10.1111/tct.70472
7. Johns Hopkins Sheridan Libraries and University Museums. AI foundations for academic work. Updated Aug. 24, 2026. Accessed Aug. 21, 2026. https://guides.library.jhu.edu/ai-foundations
8. Ke Y, Jin L, Ong JCL, Thirunavukarasu AJ. AI-induced never-skilling in medical education. Nat Med. 2026;32(6):1997-2006.
9. Komasawa N, Yokohira M. Generative artificial intelligence (AI) in medical education: a narrative review of the challenges and possibilities for future professionalism. Cureus. 2025;17(6). doi:doi:10.7759/cureus.86316
10. Lama A. Why Professional Identity Formation Matters Now for Medical Educators. Harvard Medical School Professional, Corporate, and Continuing Education. Published Jan. 20, 2026. https://learn.hms.harvard.edu/insights/all-insights/why-professional-identity-formation-matters-now…
11. Lane M, Williams M. Defining and Classifying AI in the Workplace. OECD Social, Employment and Migration Working Papers, No. 290. OECD Publishing; 2023.
12. Benjamin M. The value of consensus. In: Bulger RE, Meyer Bobby E, Fineberg HV, eds. Society's Choices: Social and Ethical Decision Making in Biomedicine. National Academies Press; 1995.
13. Myers T. Mosby's Medical Dictionary. 8th ed. Mosby Elsevier; 2009.
14. National Academies of Sciences, Engineering, and Medicine. Test and Evaluation Challenges in Artificial Intelligence-Enabled Systems for the Department of the Air Force. The National Academies Press; 2023. https://doi.org/10.17226/27092
15. Nguyen A, Ngo HN, Hong Y, Dang B, Nguyen BPT.  Ethical principles for artificial intelligence in education. Educ Inf Technol. 2023;28(4):4221-4241. https://doi.org/10.1007/s10639-022-11316-w
16. Organisation for Economic Co-operation and Development (OECD). Explanatory Memorandum on the Updated OECD Definition of an AI System. OECD Artificial Intelligence Papers, No. 8. OECD Publishing; 2024. https://doi.org/10.1787/623da898-en
17. Ouyang F, Jiao P. Artificial intelligence in education: the three paradigms. Computers and Education: Artificial Intelligence. 2021;100020(2). https://doi.org/10.1016/j.caeai.2021.100020
18. Pope Leo XIV. Magnifica humanitas: on safeguarding the human person in the time of artificial intelligence. The Holy See. Published May 15, 2026. https://www.vatican.va/content/leo-xiv/en/encyclicals/documents/20260515-magnifica-humanitas.html
19. Saeidnia HR, Jahani S, Ghiasi N, Keshavarz H. Generative AI and health misinformation: production, propagation, and mitigation-a systematic review. BMC Public Health. 2026;26(1):693. https://doi.org/10.1186/s12889-025-26148-9
20. Schon DA. The Reflective Practitioner: How Professionals Think in Action. Basic Books; 1984.
21. Tyrangiel J. AI for Good: How Real People Are Using Artificial Intelligence to Fix Things That Matter. Simon & Schuster; 2026.
22. U.S. Food and Drug Administration. Digital health and artificial intelligence glossary – educational resource. Updated Sept. 26, 2024. https://www.fda.gov/science-research/artificial-intelligence-and-medical-products/fda-digital-healt…
23. U.S. Government Publishing Office. National artificial intelligence initiative. Published Jan. 1, 2021. www.govinfo.gov: https://www.govinfo.gov/content/pkg/USCODE-2020-title15/html/USCODE-2020-title15-chap119.htm
24. U.S. Department of Justice. Community Relations Services Toolkit for Policing: Understanding Bias: A Resource Guide. U.S. Department of Justice Community Relations Service; 2015. https://www.justice.gov/d9/fieldable-panel-panes/basic-panes/attachments/2021/09/29/understanding_b…
25. Vassilev A, Oprea A, Fordyce A, Anderson H. Adversarial Machine Learning: A Taxonomy and Terminology of Attacks and Mitigations. National Institute of Standards and Technology and U.S. Department of Commerce; 2024.
26. Braveman P. What are health disparities and health equity? We need to be clear. Public Health Rep. 2014;129(1):5-8.