The Featured Plenary Harnessing AI Agents in the GME Office: Innovations and Future Directions, drew a full room of engaged attendees eager to explore how artificial intelligence (AI) is already reshaping graduate medical education (GME), and what lies ahead. Led by ACGME Chief Information Officer Bruce Metz, PhD, CHCIO and featuring a panel of experienced innovators, the session delivered a thoughtful, grounded, and forward‑looking discussion on the promise and responsibility of AI in GME.
AI Is Inevitable; Intentionality Is the Choice
Dr. Metz opened the session with a clear and pragmatic message: AI is here whether organizations are ready or not. The real choice, he emphasized, is whether institutions move forward reactively or with intention and strategy. Rather than viewing AI as a threat, Dr. Metz encouraged attendees to consider how AI agents can expand organizational impact while preserving, and even strengthening, humanity in GME.
He introduced “The Bridge” as a guiding metaphor, inviting GME leaders to cross thoughtfully into AI adoption. When approached responsibly and productively, he said, AI agents have the potential to be truly game‑changing and to enhance creativity, drive innovation, and free staff members from repetitive, time‑consuming tasks so they can focus on higher‑value work.
Balancing Opportunity, Risk, and Oversight
Dr. Metz outlined the most pressing opportunities and risks associated with AI agents, underscoring the importance of human oversight throughout every stage of adoption. He stressed that AI implementation should be staged, governed, and supported by strong guardrails to ensure safety, ethics, and accountability.
The ultimate goal, he noted, is not replacement but elevation: allowing people to spend more time on judgment, learning, and care, areas where human expertise and empathy matter most.
To ground the conversation in reality, Dr. Metz shared a word cloud generated from community surveys capturing current sentiments about AI. Prominent themes included ethics, concerns, potential, and implementation in medicine. The session was intentionally designed to address these perspectives head‑on, acknowledging both excitement and hesitation within the GME community.
Practical Paths to Implementation
Woodson Jones, MD, vice dean for GME, designated institutional official (DIO), and professor of Pediatrics at the University of Texas Health Sciences Center at San Antonio, built on this foundation by discussing commercial, off‑the‑shelf enterprise AI solutions available to institutions today. Drawing from extensive experience planning and rolling out AI systems, he provided a broad perspective on how GME stakeholders can implement tools that optimize workflows and outcomes while mitigating risk. His insights highlighted the importance of selecting scalable, secure solutions aligned with institutional needs and values.
He closed his portion noting that his AI GME dreams include “reducing drudgery, enhancing efficiency, and allowing more time for creativity!”
Hadar Duman, MHI, director of Accreditation at the University of Washington (UW) School of Medicine, shared her institution’s GME AI landscape, offering a compelling look at how AI is already being used in educational environments. UW’s AI agents assist with reviewing and summarizing materials and generating drafts for staff to refine, she said. In a live demonstration, Ms. Duman showed how tasks that once took 45 minutes to an hour can now be completed in about 60 seconds. This transformation reduced labor from eight dedicated team members to two, cut more than 100 hours of work down to 40, and improved consistency across processes. While highlighting these gains, her stance was clear that "human accountability and manual oversight remain essential to ensure quality, accuracy, and legal compliance."
AI in Accreditation, Oversight, and Program Improvement
Jie Li, PhD, senior program manager for GME at Stanford Health Care, focused on AI‑powered automation for ACGME Accreditation Data System updates, an area known for its complexity and heavy staff burden. She explained how standards‑based AI approaches can enhance oversight and accreditation by handling document review more efficiently and consistently. A video demonstration showed an AI agent reviewing documents at a fraction of the time required for human review, resulting in a 64 percent reduction in labor while producing cleaner, more standardized outputs.
Finally, Pedro Tanaka, MD, PhD, MACM, associate DIO and associate program director at Stanford Health Care, closed the panel by introducing Stanford’s Program Evaluation & Synthesis Framework. This approach leverages AI‑assisted qualitative analysis to identify key strengths and areas for improvement across programs. By synthesizing large volumes of feedback into a formulated action plan, PRISM reduced analysis timelines from weeks to hours, while improving consistency, scalability, and objectivity. The results included measurable improvements in program experience and satisfaction.
Shared Themes and Lasting Takeaways
Throughout the session, audience members remained visibly and audibly engaged, taking notes and responding enthusiastically to real‑world examples. Several concise takeaways from the panelists captured the spirit of the discussion:
- Woodson Jones: “AI is a tool that is inevitable. Responsibility over speed will free, not fail, people.”
- Hadar Duman: “AI is a toddler. We have to engage now to shape how it turns out as an adult.”
- Jie Li: “People should allow AI to handle the robotic work so we can focus on being humans.”
- Pedro Tanaka: “We need to be fully engaged in the AI implementation process.”
The session closed by returning to “The Bridge” metaphor, reinforcing a call to action for GME leaders: Be proactive, intentional, and collaborative in AI adoption. Embracing this shared AI journey offers a powerful opportunity to reimagine how GME offices can work more efficiently, more thoughtfully, and more human‑centered than ever before.