AI in Healthcare Series: You Can’t Delegate AI with Andy Slavitt
TL;DR
Andy Slavitt and Stanford experts discuss how AI adoption in healthcare is accelerating faster than organizational readiness, creating a critical imperative for leaders to personally engage with the technology while new CMS payment models and FDA pathways signal a policy shift toward rapid AI integration and patient empowerment.
🎯 The "Don't Delegate AI" Mandate 2 insights
Leaders must personally engage with evolving tools
Slavitt warns that AI capabilities change too rapidly for executives to delegate understanding to IT teams; leaders dismissing AI based on two-year-old experiences with ChatGPT hallucinations are making decisions on obsolete technology.
Hire dedicated talent to maintain currency
Town Hall Ventures hired a full-time AI research fellow solely to keep the investment team at the cutting edge, acknowledging they "can't sound smart at something they read about" and must use the tools daily to evaluate opportunities.
⚠️ The Healthcare Adoption Disconnect 2 insights
Physicians adopt AI despite organizational barriers
Survey data reveals 67% of physicians use AI daily, yet 81% are dissatisfied with their organization's approach, with 71% having no influence over tool selection and 48% rating employer AI communication as poor.
Bottom-up pressure meets restrictive policies
Unlike previous healthcare technologies requiring change management "push," clinicians are "pulling" AI through personal devices, creating a "red alert" situation where health systems risk irrelevance by restricting rather than empowering usage.
🏛️ Policy Signals and Payment Innovation 2 insights
CMS Access Model shifts payment to technology outcomes
The new CMS model pays technology companies—not just physicians—when AI tools demonstrate measurable improvements in patient health, representing the strongest policy signal yet for rapid deployment and outcomes-based validation.
FDA creates fast-track approval pathway
The FDA established a separate review process allowing AI tools to bypass traditional long-term safety studies, while the administration actively promotes adoption through Treasury dollars rather than merely preventing state-level regulations.
👥 Patient Empowerment and Access Transformation 2 insights
Consumers rapidly closing the information gap
Comfort with "AI doctors" jumped from 11% to 28% in one year, with patients creating personal health LLMs containing complete lab histories to prepare for visits, fundamentally reducing traditional physician-patient information asymmetry.
Rural access becomes transformative necessity
For rural patients facing nine-week specialist waits, AI represents essential infrastructure rather than convenience, providing 24/7 access to guidance that exceeds traditional physician text relationships and addresses critical care gaps.
Bottom Line
Healthcare leaders must immediately shift from trying to 'tame' AI through restrictive policies to actively 'riding' it by personally using the tools, empowering clinical adoption, and preparing for patients who arrive with AI-generated health insights.
More from Stanford Online
View all
Stanford CS221 | Autumn 2025 | Lecture 20: Fireside Chat, Conclusion
Percy Liang reflects on AI's transformation from academic curiosity to global infrastructure, debunking sci-fi misconceptions about capabilities while arguing that academia's role in long-term research and critical evaluation remains essential as the job market shifts away from traditional entry-level software engineering.
Stanford CS221 | Autumn 2025 | Lecture 19: AI Supply Chains
This lecture examines AI's economic impact through the lens of supply chains and organizational strategy, demonstrating why understanding compute monopolies, labor market shifts, and corporate decision-making is as critical as tracking algorithmic capabilities.
Stanford CS221 | Autumn 2025 | Lecture 18: AI & Society
This lecture argues that AI developers bear unique ethical responsibility for societal outcomes, framing AI as a dual-use technology that requires active steering toward beneficial applications while preventing misuse and accidental harms through rigorous auditing and an ecosystem-aware approach.
Stanford CS221 | Autumn 2025 | Lecture 17: Language Models
This lecture introduces modern language models as industrial-scale systems requiring millions of dollars and trillions of tokens to train, explaining their fundamental operation as auto-regressive next-token predictors that encode language structure through massive statistical modeling.