At the frontline of patient care, providers have been put under impossible pressure to lead the charge for AI within healthcare. This is not a sustainable way to move the needle towards increased AI innovation for healthcare as a whole. Providers require a familiar frame of reference, collaboration with a broader network to build effective standards, and tools that will help translate the impact of innovation to patients and ensure safety.
The healthcare industry is finally on the brink of a transformative shift with AI to accomplish this, after years of unrealized potential and broken promises. In July, the White House introduced America’s AI Action Plan to accelerate AI adoption. In particular, the plan cites critical sectors, like healthcare, as being especially slow to adopt due to a variety of factors, including distrust or lack of understanding of the technology, a complex regulatory landscape, and a lack of clear governance and risk mitigation standards. This plan reflects a decisive shift: one that prioritizes speed, cost-effectiveness, and innovation while preserving necessary safeguards. Its overarching message is clear – AI must move faster, smarter, and with leadership accountability and cross-industry collaboration. We agree.
The key to actually achieving this lies in leveraging familiar frames of reference for both patients and providers that are built on a proven track record for success.
Building Upon a Successful Foundation
Healthcare is already one of the most highly regulated industries. Rather than imposing entirely new regulatory structures, a more practical approach is to determine how existing oversight frameworks – offered by agencies like the FDA and the Office of the National Coordinator for Health Information Technology (ONC), among others— can be applied to guide the responsible use of AI in healthcare. These current standards are well-suited to begin testing and implementation in lower-risk, administrative AI applications such as clinical documentation automation, billing support, and memo generation, as they enhance efficiency and reduce costs without introducing significant clinical risk. This will help increase adoption by providers across the nation while ensuring a powerful threshold for safety. From there, they can potentially move to evaluate high risk clinical algorithms.
Other successes to build upon exist within federal healthcare institutions, such as the U.S. Department of Veterans Affairs and the National Institutes of Health (NIH). These organizations are uniquely positioned to demonstrate leadership in responsible AI adoption by highlighting existing efforts, programs, and training initiatives to showcase clear examples of successful AI deployment thus far and help contribute to the development and validation of recommended benchmarks.
The combination of existing regulations and proven successes will encourage increased adoption while also providing an effective frame of reference for collaborative bodies that will result from the AI Action Plan.
Manage Risk: Guardrails for AI Adoption
Trust is another persistent barrier to AI adoption – especially in healthcare, where stakes are high and missteps can have life-altering consequences. Building confidence in AI tools goes beyond technical validation; it requires transparent performance metrics, clear accountability, and rigorous documentation. These qualities should be clearly communicated to help patients, doctors and all healthcare users understand an AI system’s purpose, capabilities, and limitations. The OMB’s April 2025 M-25-21 memo underscores this point by mandating government agencies to evaluate “high-impact AI” systems – systems that can affect individual rights, access to critical services, public safety, or human health. These systems must undergo enhanced risk assessments, including documentation of model assumptions, limitations, and scope of use.
In healthcare, that means an AI application – such as one used for clinical decision support or diagnostics – that impact patient outcomes should be subject to higher scrutiny and compliance thresholds before deployment. Conversely, AI applications that don’t affect patient outcomes should be deployed with less scrutiny and review. Both scenarios can be effectively evaluated by building and following a standard checklist that includes things such as secure design, continuous monitoring, bias mitigation, and robust data governance.
This would operate akin to a nutrition label where consumers expect a level of transparency as to what they’re consuming. The same clarity is warranted from the AI tools that could potentially influence patients’ health protocols, diagnosis, and treatment plans. A nutrition label would serve as a common language for evaluating AI systems. This way doctors and care teams can consistently and confidently compare applications to pick what is best for their patient population. And vendors know what characteristics and performance metrics to put forward to compete in the market.
A “nutrition label for AI” would outline intended use, model performance, training data summaries, and known limitations. These serve as “product labels” for AI applications, helping stakeholders—from clinicians to regulators—evaluate system readiness, fairness, and safety. Performance metrics should be versioned and regularly updated, and red teaming protocols must test systems for adversarial risks or misuse. The success of AI, especially in healthcare, depends on strong governance that prioritizes reliability and safety to earn public trust.
The Path Forward
The transformative potential of AI in healthcare is undeniable. However, realizing the full benefits of AI demands a disciplined and thoughtful approach. By leveraging existing regulatory frameworks, fostering cultural readiness, and promoting collaboration we can pave a responsible path for AI adoption. To get there responsibly, federal healthcare leaders must act with urgency and care, aligning with relevant parts of White House’s AI Action Plan and OMB’s standards by implementing standardized AI documentation practices and conducting rigorous pre-deployment risk assessments.
The ultimate goal is to enhance patient care while maintaining public trust and safety. Moving from theory to practice requires a collective effort to bridge the gap between technological possibility and practical, regulated application.
About Kevin Vigilante
Kevin Vigilante is the former Chief Medical Officer at Booz Allen Hamilton, where he also lead the Health Futures Group. He is currently an advisor for Booz Allen. In his former role as CMO, Kevin advised government healthcare clients at the Departments of Health and Human Services, Veterans Affairs, and the Military Health System. A physician at his core, Kevin is passionate about offering new ideas for health system planning, biomedical informatics, life sciences and research management, and public health – largely through the lens of digitally-enabled care.
About Dr. Dave Prakash MD
Dr. Dave Prakash MD, is a physician-technologist focused on AI enablement at Booz Allen Hamilton. He provides clinical expertise for health innovation and AI for public sector and commercial clients. He recently led AI governance, creating the policies, processes and infrastructure to ensure safe and responsible AI practices within the company and for its clients. Prior to Booz Allen, Dave contributed to the development of AI solutions for C3 AI and Elevance Health, where his responsibilities spanned product development, clinical consultant, business development, and government relations.
