homepage-image-2

Defining the Role of the Physician in the Digital and AI Era of Medicine

A closed-door working session hosted by the American Medical Association (AMA) and the Digital Medicine Society (DiMe)

May 21, 2026 • 10:00 am - 7:00 pm 

National Academy of Sciences Building, 2101 Constitution Avenue, NW, Washington, DC

Thanks for Attending

See photos from the event.

Photos from this event may be shared publicly, provided they include no captions, tags, or context that reference discussions or statements made during the event. In keeping with Chatham House Rule, the identities of attendees should not be disclosed. Please also obtain written consent from anyone identifiable in your photos before sharing.

See photos
AMA & DiMe_052126-280

Why this convening

AI is moving faster than consensus on the physician’s role. 

Without a clear, trusted definition, scope of practice will be set by markets and regulation instead of by the profession. 

DiMe and the AMA have synthesized the best available research on AI capabilities, physician roles, and governance. This convening will guide us to critical consensus.

AMA_1C_269
DiMe

Agenda

A single day of focused discussion with leaders from practice, technology, policy, and education

The day is structured around four discussions, each focused on a different dimension of the physician’s role in an AI-enabled system.

Download the full agenda
Session 1 – The Art of the Possible

Outcome: Establish a shared, evidence-based picture of what AI can and cannot do in medicine today, and where the biggest uncertainties lie.

Key Questions:

  1. How do we understand current AI capabilities in medicine based on demonstrated performance, and where are the biggest uncertainties?
  2. How should physicians interpret performance claims from benchmarks compared with live deployments?
Session 2 – AI in Practice

Outcome: Ground the conversation in real-world experience: how AI is already changing workflow, team roles, and patient care.

Key Questions:

  1. How is AI reshaping workflow, cognitive load, and team responsibilities in real clinical settings today?
  2. How do we distinguish between deployments that improve care and those that simply shift or add burden?
Session 3 – Defining the Physician’s Role

Outcome: Clarify what physicians must remain directly accountable for, and how they lead teams that use AI safely and effectively.

Key Questions:

  1. How should physicians define their non-delegable responsibilities as AI tools take on more diagnostic and workflow tasks?
  2. How should physicians lead teams and stay accountable when AI systems contribute to clinical decisions?
Session 4 – Policy Foundations

Outcome: Identify the payment, licensure, and incentive changes needed to support the physician role defined in Session 3.

Key Questions:

  1. How do payment and licensure structures need to change to support physicians leading AI-enabled care?
  2. How should expectations for evaluation, documentation, and oversight of AI tools be framed?
  3. How should incentives align so that failing to use well-validated AI does not inadvertently limit access to high-quality care?

 

Pre-read package

The following materials provide the shared foundation for the day’s discussion. Participants are asked to review them in advance so that the convening can focus on deliberation rather than orientation.

🔗 QUICK LINKS

AI CAPABILITIES IN MEDICINE
THE PHYSICIAN'S ROLE
GOVERNANCE, INCENTIVES & POLICY CONTEXT
CLINICAL VIGNETTES

AI capabilities in medicine

The evidence base for AI in medicine is growing rapidly, but unevenly.

Diagnostic imaging & pattern recognition

AI systems have demonstrated performance comparable to specialists in radiology, pathology, and dermatology in controlled settings. Translation to live clinical environments is less consistent, with performance often degrading outside the populations and workflows used in training.

Clinical decision support

AI-driven tools are entering workflow for sepsis prediction, medication interactions, and risk stratification. Early evidence suggests improvements in speed and consistency, though clinician override rates remain high and alert fatigue is a growing concern.

Administrative & documentation tasks

Ambient listening, auto-coding, and note generation are among the fastest-adopted AI applications. These tools reduce documentation burden but raise questions about accuracy, liability for AI-generated records, and shifts in how physicians spend their time.

Drug discovery & genomics

AI is accelerating candidate identification and trial design. These applications are further from the bedside but will reshape the therapeutic landscape physicians practice within.

Key uncertainties

Performance claims from benchmarks frequently overstate real-world results. Bias, equity, and generalizability remain unresolved. The FDA does not regulate the practice of medicine and has stated it never will unless federal law changes — regulatory oversight of AI tools and clinical accountability for their use are governed by different frameworks.

The physician’s role

How it is being debated

Across medicine, law, and policy, there is active disagreement about what should remain uniquely physician-led as AI takes on more diagnostic, predictive, and administrative functions.

DiMe-AI-gov

Ultimate oversight and care of the patient remains a broadly accepted physician responsibility — but what “oversight” means when AI mediates more decisions is contested.

Leadership of the care team is evolving as AI enables other team members to operate at higher levels of autonomy.

The framing question is not which specific tasks physicians will keep, but what should remain true about the physician’s role as technology evolves — a principles-based approach rather than a task-based one.

Technology within the existing framework

View: AI will be adopted, but the care delivery model and physician role remain largely unchanged.

Implication: Focus on guardrails, training, and integration within current workflows.

Technology redefining the framework

View: AI will significantly reshape what physicians do, how teams function, and what accountability looks like.

Implication: Focus on redefining scope, redesigning teams, and updating regulatory and payment structures.

Governance, incentives, and policy context

Payment models

Payment models

Current CMS reimbursement is tied to billable minutes and face-to-face encounters. This structure may not support a physician role that shifts toward oversight, team leadership, and AI-mediated decision-making.

Remember:

  • Bullets are great
  • For spelling out benefits and
  • Turning visitors into leads.
State medical boards

State medical boards

Certification and continuing education requirements are not yet aligned with the competencies AI-enabled practice demands. Licensure frameworks vary by state and have not kept pace.

Highlights

  • Bullets are great
  • For spelling out benefits and
  • Turning visitors into leads.
Incentive alignment

Incentive alignment

If well-validated AI tools improve outcomes but their use is neither required nor reimbursed, adoption will be uneven and access disparities may widen. Conversely, failing to use validated AI could become a standard-of-care question.

Highlights

  • Bullets are great
  • For spelling out benefits and
  • Turning visitors into leads.

Clinical vignettes

The following scenarios anchor the day’s discussion in concrete situations.

radiologist

VIGNETTE 1

AI-assisted diagnosis in radiology

A radiologist reviews 80 chest CTs per shift. An AI tool flags three studies as high-probability lung nodules, two of which the radiologist agrees with and one she does not. The AI’s sensitivity exceeds hers on published benchmarks. She overrides the AI. The patient is later diagnosed with early-stage cancer at a different facility.

Who is accountable? Does the benchmark data change the standard of care?

VIGNETTE 2

AI-driven team workflow in primary care.

A primary care physician oversees a panel of 2,400 patients. An AI system triages inbox messages, pre-drafts responses for the care team, and flags patients for proactive outreach based on risk scores. The physician reviews and signs off on AI-recommended actions but has not personally assessed most of the flagged patients.

Is this delegation or abdication? What does meaningful oversight look like at this scale?

clinician
clinician-2

VIGNETTE 3

Payment and incentives misalignment.

A health system deploys an AI-powered sepsis prediction tool that reduces ICU admissions by 15%. The physicians who championed it spent significant time on validation, workflow redesign, and team training. None of this work is billable. Meanwhile, the reduction in ICU days lowers revenue under the current fee-for-service model.

How should the physician’s role in AI deployment be recognized and compensated?

Suggested reading

  1. [Placeholder – Selected journal article on AI diagnostic performance and real-world generalizability]
  2. [Placeholder – Selected perspective on physician accountability in AI-mediated care]
  3. [Placeholder – Selected policy analysis on CMS payment models and AI-enabled care delivery]
  4. [Placeholder – Selected commentary on the evolving physician-team relationship in AI-enabled settings]

Confirmed speakers

Session chairs & moderators

Jen_Goldsack_headshot

Jennifer Goldsack

CEO | Digital Medicine Society (DiMe)

Co-host

LinkedIn
Aelaf _Worku

Aelaf Worku

VP, National Medical Director | SCAN Health Plan

Speaker

LinkedIn
Alexander Billioux

Alexander Billioux

Chief Health Officer | Cityblock Health

Speaker

LinkedIn
Amy Compton-Phillips

Amy Compton-Phillips

CMO | CVS Health

Speaker


LinkedIn
Anna M.

Anna McCollister

Patient Expert 

Speaker


LinkedIn
Ateev Mehrotra

Ateev Mehrotra

Professor of Health Services | Policy & Practice, Brown University

Speaker

LinkedIn
Ben-Vandendriessche-headshot

Ben Vandendriessche

President and Chief Scientific Officer | Digital Medicine Society (DiMe)

Moderator

LinkedIn
Bobby Mukkamala

Bobby Mukkamala

President | AMA

Moderator


Info
Humayun Chaudhry

Humayun Chaudhry

President & CEO | Federation of State Medical Boards (FSMB) 

Speaker

LinkedIn
Jacob Shiff

Jacob Shiff

Chief AI & Technology Officer | CMS Innovation Center

Speaker

LinkedIn
James H

James Hairston

Global Head of Innovation Policy | OpenAI

Speaker

LinkedIn
Kurt-Herzer-image

Kurt Herzer

Head of Healthcare & Life Sciences Policy | Anthropic

Speaker

LinkedIn
Matt Diamond

Matt Diamond

CMO, Digital Center of Excellence | FDA

Speaker

LinkedIn
Matt Pavelle

Matt Pavelle

Co-Founder & CEO | Doctronic

Speaker

LinkedIn
Pete Clardy

Pete Clardy

Director, Clinical Enterprise | Google for Health

Speaker

LinkedIn
Priyanka Agarwal

Priyanka Agarwal

Co-Founder & CEO | HealthEx

Speaker

LinkedIn
Reid Conant

Reid Conant

Sr. Physician Executive | Abridge

Speaker

LinkedIn
Ryan Vega

Ryan Vega

Chief Health Officer | Vantiq

Moderator

LinkedIn
Susan Sly

Susan Sly

Founder and CEO | The Pause Technologies Inc.

Speaker

LinkedIn
Toju Duke

Toju Duke

Founder | Diverse AI

Speaker

LinkedIn

Pre-read package

Come prepared.

The following resources are intended to provide an evidence-based foundation for our convening. Participants are encouraged to engage with the priority reads prior to our meeting on May 21, while the supplemental materials offer additional depth for those seeking a more comprehensive perspective.

Priority reads

From Benchmark to Bedside: The Current State of AI in Clinical Practice

The evidence base for this convening. Synthesizes what AI can and can't do in clinical care today, and frames the core questions we aim to address.

From Promise to Practice: The Next Era of AI in Health Care

Provides a high-level overview of the essential conditions, spanning reimbursement, policy, and technical domains, required to ensure the safe and durable expansion of medical AI.

Additional reads

State of Clinical AI Report 2026
Stanford overview
Analyzes the gap between theoretical AI performance and clinical reality, offering a candid assessment of operational friction and implementation challenges.

From Agents to Governance: Essential AI Skills for Clinicians
Outlines a tripartite framework for clinician skills, detailing the evolution of the physician from a recipient of digital data to an essential overseer of autonomous systems.

Patients are not waiting for permission (Lancet Comment)
Presents a direct argument regarding autonomous patient use of AI outside traditional clinical paths, repositioning the clinician as a strategic partner in interpretation instead of a solitary gatekeeper.

AMA Augmented Intelligence Development, Deployment, and Use
Outlines foundational AMA standards for safety and algorithmic transparency, serving as the essential policy framework for our discussion. 

 

Logistics

Date: May 21, 2026
Time: 10:00 am - 7:00 pm
Location: National Academy of Sciences Building, 2101 Constitution Avenue, NW, Washington, DC 
📍 View map

In-person, invitation-only working session

Getting to the National Academy of Sciences (NAS) Building

  • Address: 2101 Constitution Avenue NW, Washington, DC 20418
  • Entry: All guests should enter through the C Street doors.
Hotels

While there is no official conference hotel, the following hotels are convenient options within walking distance of the meeting site at the National Academy of Sciences Building:

  • State Plaza Hotel | 2117 E Street NW
    (202) 861-8200
    ~0.2 miles (about a 5-minute walk)

  • Hotel Lombardy | 2019 Pennsylvania Avenue NW
    (202) 828-2600
    ~0.7 miles (about a 10–15 minute walk)

  • The River Inn | 924 25th Street NW
    (202) 337-7600
    ~0.7–1 mile (about a 15-minute walk)

Airports

The meeting site is approximately 5 miles from Washington National Airport (a 20-minute cab ride depending on the time of day) and approximately 25 miles from Dulles International Airport (a 45-minute cab ride).

Metro

The Foggy Bottom metro stop (Orange/Blue Line) is located at 23rd and I Streets NW. Walking from the metro to the NAS building takes approximately 12 minutes.

Parking

The National Academy of Sciences parking garage is located on 21st Street NW between Constitution Avenue and C Street. Parking is limited; alternative transportation is recommended.

Taxis & Rideshare

Drop-off and pickup are on 22nd Street, at a shared taxi stand with the State Department. From there, walk along C Street to the entrance, about midway down the block.

Alyssa-headshot

Contact us

Contact Alyssa Cummings, Events Lead, at alyssa@dimesociety.org for logistical questions.

Let the countdown begin!

The countdown is complete

How this fits into AMA’s broader work

This convening is one step in a larger effort to define and support the physician’s role as medicine evolves.

Evidence

DiMe has synthesized the best available research on AI capabilities and physician roles.

This convening

Senior leaders test that picture against reality and define what physicians should own as AI evolves.

Follow-through

AMA translates the discussion into guidance and tools for physicians, health systems, and policymakers.

Organizing partners

Why this convening

Grounding in evidence

DiMe has synthesized the best available research on AI capabilities, physician roles, and governance.

This convening

A single day of focused discussion with leaders from practice, technology, policy, and education.

What follows

AMA guidance and tools that help physicians and health systems lead AI-enabled care.