Recently, while U.S. tech leaders united in a call for a pause in AI development, Anthropic CEO Dario Amodei also indicated that healthcare could be an exception – speculating that AI could help cure most diseases within five to ten years.
Against this wider debate over whether AI is moving too fast, I asked a wide array of healthcare, biotech, AI and finance leaders in the U.S. a more immediate question: Should AI in healthcare slow down? Or should we wait for someone to say stop? Or should we hit the gas and accelerate? Whatever their views on the pace of AI development, participants agreed that in health, AI must accelerate. It is essential to the present and the future of care.
To get there, we must be clear-eyed about what AI can actually do in medicine. It will never replace a doctor, and while it shows tremendous promise in the quest to more rapidly cure disease, the health and AI leaders I spoke with were skeptical it can do so in the next 5-10 years.
We aren’t there yet; that distinction matters. Discovering a potential treatment is only the beginning of a long journey through testing, clinical trials and delivery to patients. AI may and must accelerate that journey. Though we cannot promise it will cure most diseases on an accelerated timetable, we can already see its revolutionary impact on patients now.
AI is already helping clinicians detect disease earlier, making complex medical information easier for patients to understand, and reducing time-consuming administrative work for physicians. It is also helping hospitals anticipate demand, manage patient flow and make better use of scarce beds, operating rooms and clinical expertise.
These gains may sound modest – banal even — compared with the promise of curing cancer. But they contribute to life-altering outcomes for patients waiting for a diagnosis, a hospital bed or an appointment with a specialist.
At King Faisal Specialist Hospital and Research Centre, an AI-enabled command center monitors patient flow, anticipates bottlenecks and alerts teams when intervention is needed. The technology became useful when we paired its data with the experience of nurses and other staff who knew why delays occurred. Together, we changed admission and discharge processes and clarified who was responsible for resolving problems. Emergency room bed waiting times fell from 32 hours to six.
AI helped us see the problem. Our people helped design the solution. Patients experienced the benefit. This is the new teamwork model healthcare must build – clinician and AI working together for the patient.
This is why technology leadership cannot remain a separate obsession at the edge of a healthcare organization. In our case, we have elevated our chief information officer to stand alongside clinical department leaders making decisions at our hospital. Technology teams need to feel the same urgency clinicians feel when looking into the eyes of a patient waiting for care. Clinical teams, in turn, must help determine which problems technology should solve.
As a physician who now leads a hospital, I want AI to help discover cures. But patients also deserve better care today. Much of the focus has been on proving AI can help with routine administrative tasks. The next test is moving from successful pilots to greater scale – bringing the benefits of AI to the broader system and expanding clinical use cases.
Scaling AI requires building trust within the patient-clinician-AI relationship. Clinicians need to understand the support AI can bring to their treatment of patients, and patients need to understand that AI can support better health outcomes, including earlier diagnoses and more dedicated time with experts. We must also be deliberate about what we ask AI to optimize. In clinical care, AI tools should go beyond financial and operational efficiencies — they must be trained to put the patient at the center of all decisions.
This is where AI in health must accelerate: identifying what matters most for patients, testing tools in real clinical settings, and designing workflows to scale tools that work. Data is integral to this process. Better and more representative data makes AI more capable and safer for diverse patient groups. Yet health data remains fragmented across institutions, incompatible systems, and countries.
Individual health systems do not have enough information to represent every population they hope to serve. But with strong privacy and cybersecurity standards, responsible collaboration can allow health systems to learn from one another to better serve their patients.
This matters if AI is to democratize healthcare. Every year that reliable, proven tools and clinical knowledge cannot cross institutional or national boundaries is another year in which expertise remains available only to patients fortunate enough to live near it.
Amodei’s concern about the risks of powerful AI is worth discussion. His ambition for AI in health is welcome. As healthcare leaders, we must remember to also account for what patients may lose when progress moves too slowly. Patient outcomes must come first.
We may disagree on how soon AI will cure disease. But it is already helping solve problems patients face now. A responsible approach must protect us from the consequences of moving too quickly without ignoring the human cost of moving too slowly.
