AI Alone Won’t Transform U.S. Healthcare

Imagine it’s the middle of the night, and your child is seriously ill. You rush to the hospital, bracing for mountains of paperwork, long waits, and an overburdened staff. Instead, you meet a clinician who knows your child’s medical history in precise detail. Within minutes, a specialist team evaluates your child and prescribes a personalized treatment plan tailored to her genetics and thousands of similar cases. No guesswork. No redundant testing. The diagnosis is clear, as is the path forward. Artificial intelligence could make this story the new normal. Not since the discovery of antibiotics has medicine faced such transformative potential. But AI alone will not deliver this future without serious structural reforms in the United States. We need to fix the flaws that have broken the current healthcare system. If we don’t, AI could amplify the inefficiencies, misaligned incentives, and inequities that pervade healthcare today. With AI evolving at a blinding speed, the window to act is closing—along with the opportunity to ensure AI’s benefits reach everyone. At Stanford, where I work, and across the industry, we are already seeing AI’s real-world promise. Complex biological processes that once took scientists decades to describe are being illuminated by AI in a matter of months. AI is streamlining clinical trials, helping scientists identify new drug targets, and is even aiding in the development of novel therapeutics. The open question is whether these breakthroughs will translate into meaningful improvements in human health. Turning discoveries into better patient outcomes demands more than technological progress; it will require deep reforms in the U.S. healthcare infrastructure, payment systems, and clinical workflows. Without these changes, even the most promising innovations risk remaining stuck in research labs or limited to only those who can afford them. Avoid a Repeat of the EHR Experience The Health Information Technology for Economic and Clinical Health (HITECH) Act of 2009, which mandated the adoption of electronic health records (EHR) systems, ushered in the digital age of medicine in the United States. But rather than improving the experience of care, for years, the transition saddled clinicians with overwhelming administrative burdens. Designed largely to satisfy insurers’ complex billing requirements, EHRs often neglected the practical needs of doctors and patients. Even as EHRs have improved over time, they remain cumbersome, inefficient, and, in some cases, pose safety risks due to their user-unfriendly interfaces. Studies published in recent years have found that physicians spend a huge amount of their time on tasks related to EHRs, contributing to high burnout rates. We risk repeating this mistake with AI. Already, we see an AI arms race developing between physicians and insurance companies. Insurers are deploying AI tools to scrutinize and often rapidly deny treatment coverage, while physicians counter with AI-driven systems to justify necessary care. This technological tug-of-war doesn’t improve care delivery; it merely shifts resources from patient care to battling within the system. To steer AI toward delivering a healthier future, we must pursue three urgent priorities: 1. Revamp incentives. The first step is to confront the incentives that shape healthcare today. Historically, market incentives have profoundly influenced the adoption of technologies, and not always for the greatest good. We must carefully design incentives to promote the outcomes we want and mitigate any adverse effects. AI cannot transform a system built on misaligned incentives. Today, fee-for-service reimbursement remains the dominant payment model in the United States, rewarding the volume of care delivered, not the quality or long-term impact of that care. This model prioritizes billing over clinical impact and continues to discourage interventions that emphasize prevention, early detection, and long-term health improvement. For instance, better management of early-stage diabetes could reduce all-cause mortality in patients by nearly 20%, yet reimbursement structures in the United States often fail to support the necessary long-term interventions. And across the board, tools that enable continuous monitoring, early diagnosis, or personalized treatment planning are often unfunded and underused. The Centers for Medicare and Medicaid Services (CMS) has been steadily advancing the shift toward value-based care—championing models that prioritize health outcomes over volume. While this represents an important step in the right direction, to realize the full potential of these reforms, we must ensure that AI is embedded in the next generation of value-based care strategies. That includes updating billing codes to reflect AI-enabled procedures, expanding coverage for digital diagnostics and remote monitoring tools, and creating reimbursement pathways that reward the use of AI in preventing disease, not just treating it. Legislation like the recently introduced bipartisan Health Tech Investment Act offers a timely and promising framework. The bill proposes expanding reimbursement pathways for digital health technologies, including AI, and enabling their adoption across diverse clinical settings. Structured financial support like this is critical to avoiding a fragmented future where only well-resourced institutions have a role in shaping AI-supported care. AI can also be part of the solution to reimbursement challenges, but it must be approached thoughtfully. By carefully deploying AI to automate processes like eligibility verification, claim submissions, and real-time error detection, we can reduce administrative burdens without exacerbating the tug-of-war between payers and providers. Done well, this could free up time and resources to focus on what matters most: the patient. 2. Embed AI into medical education and training. EHRs fundamentally changed the way healthcare providers work, often in ways that weren’t fully anticipated or understood. Studies on EHR rollouts highlighted significant workflow disruptions, unsafe workarounds, and documentation errors, often due to gaps in training and understanding. AI will bring equally profound changes, creating a significant training challenge and cultural shift. In medicine, we contend with a significant “know-do gap”: the lag between when we know something is beneficial and when it is actually adopted in practice. On average, it takes 17 years for new evidence-based practices to become the standard practice in the United States, largely because of a stubborn commitment to entrenched practices and the daunting volume of emerging data that healthcare providers must interpret. Overcoming these barriers requires continuous education and systemic changes to encourage rapid integration and adaptation of new knowledge. Integrating AI into medical education curricula is a critical part of the solution. Future generations of providers must be educated and equipped with the skills and mindset necessary to integrate AI into patient care. Beyond the classroom, health systems must also expand training opportunities and foster a culture of continuous learning to ensure that their healthcare professionals are ready to embrace AI as a tool to enhance what they do best. The opportunities are, indeed, significant. Consider the use of AI-supported tumor boards, which could revolutionize oncology care. Tumor boards are specialized, multidisciplinary teams of healthcare professionals who meet to discuss cancer cases and determine the best treatment plans for patients. AI could enhance how tumor boards assimilate the latest research, patient data, and expert insights, helping close the know-do gap in implementing new cancer treatments. With research grant support, Stanford Medicine is actively exploring this type of AI integration. 3. Engage doctors and patients in AI development and use. Because the development of EHRs largely excluded the voices of doctors and patients, the tools didn’t meet their needs. Recent surveys show that 50% of physicians feel EHRs have compromised their clinical effectiveness, and nearly 60% advocate for a complete overhaul of these systems. For AI to avoid this pitfall, physicians and patients must be actively involved in the development, testing, and implementation of AI tools from the beginning. According to an American Medical Association (AMA) survey, 78% of physicians want clear explanations of how AI systems make decisions. Yet, many systems still lack the transparency needed for clinicians to understand and trust AI-generated recommendations—a challenge we cannot afford to overlook. If patients are to trust AI, they, too, need to be engaged. Already, there is concerning evidence of patients growing skeptical of AI. A 2023 Pew Research Center study found that 60% of Americans would feel uncomfortable if their healthcare provider relied on AI for diagnosis and treatment recommendations. To mitigate these risks, AI tools must be developed using diverse data sets that represent a broad range of populations. In addition, involving patients and clinicians in the design and testing phases will ensure that AI tools are equitable, trustworthy, and effective across different demographic groups. In fact, it’s the human factor—equipping patients and physicians to use AI to make better care decisions—that will allow for a smoother integration of these tools. As AI systems become more advanced, a critical question is how they should be used in clinical decision-making: as tools to assist physicians or as autonomous agents making decisions independently. At Stanford, we share the view with many others that in the near term AI should be used to support, not replace, clinical judgment. Over time, however, as AI models become more capable and are validated in real-world settings, the line between support and autonomy may blur. This evolution underscores the importance of rigorous, ongoing oversight and regulation. Regulatory frameworks, such as the Food and Drug Administration’s approach to monitoring continuously learning algorithms, will need to keep pace with technological change while ensuring safety, transparency, and equity. AI represents a rare opportunity to transform the future of healthcare. But this transformation will not happen automatically. We must act deliberately and learn from past mistakes. By aligning incentives, fostering a culture of continuous learning, and involving all stakeholders, we can ensure that AI delivers a better healthcare system in the United States—one that works for everyone.

AI Alone Won’t Transform U.S. Healthcare

Summary.

Partner Center