As artificial intelligence takes on more work inside hospitals, a growing number of healthcare leaders are drawing a firm line. The tools can help — but only if the people using them already know how to do the job themselves. At The Brooklyn Hospital Center in New York, that principle has become an explicit institutional requirement. Sam Amirfar, MD, the hospital’s senior vice president and chief medical and information officer, has made it a condition of AI adoption across the organization: before anyone uses an AI tool, they must be able to perform the task without it.
The Brooklyn Hospital Center’s AI Philosophy
AI Already Doing Real Work
Brooklyn Hospital Center has moved well past the pilot stage with AI. The technology already handles the kind of routine but time-consuming tasks that once occupied large portions of an executive’s day — processing reports, surfacing patterns in spreadsheets and managing administrative workflows. Furthermore, the hospital is pushing AI deeper into clinical territory. Use cases include flagging secondary diagnoses for physicians and documenting patient visits. The scope of AI deployment is expanding. However, the governing principle stays constant throughout: human competence comes first.
Dr. Amirfar’s Non-Negotiable Condition
Dr. Amirfar has made foundational knowledge a prerequisite for AI use — not a recommendation. Every staff member who uses an AI tool must independently understand the underlying task. This applies equally to seasoned clinical veterans and new hires. The condition exists because AI at Brooklyn Hospital Center is designed to augment capability, not substitute for it. Moreover, in a resource-strained community hospital environment, productivity gains from AI carry real value — but only when they build on genuine competence rather than replace it.
The Calculator Analogy That Shapes Hospital Policy
Learning Multiplication Before Using the Calculator
Dr. Amirfar frames the issue through a straightforward analogy drawn from his own household. His children do not use calculators in class. They learn multiplication and division the traditional way first. Only once they thoroughly understand those operations does a calculator become appropriate — because then it simply saves time rather than masking a gap in understanding. The same logic governs AI adoption at Brooklyn Hospital. Introducing a large language model to someone who does not understand the underlying task creates a false productivity gain. Speed without comprehension produces outputs that cannot be defended, verified or discussed intelligently.
Why the Analogy Translates Directly to Healthcare
In healthcare, the stakes are far higher than arithmetic. Clinical decisions affect patient safety and outcomes. Consequently, the cost of misplaced confidence in AI outputs is not a wrong answer on a test — it is a missed diagnosis, an inappropriate recommendation or a clinical error. Dr. Amirfar recognizes that AI tools are useful precisely because they compress work that would otherwise take hours. However, that speed is only a benefit to someone who already understands what they are looking at. For someone without foundational knowledge and experience in healthcare delivery, acceleration simply moves faster toward the wrong conclusion.
AI in Clinical and Administrative Workflows
Brooklyn Hospital Center uses AI across both administrative and clinical functions. On the administrative side, AI processes large volumes of data that would otherwise demand significant staff time. On the clinical side, the hospital deploys AI to assist physicians with identifying secondary diagnoses — a task where AI can surface patterns across a patient’s chart that a busy clinician might otherwise miss. Additionally, AI documentation tools reduce the burden of patient visit records. Together, these applications reflect a deliberate strategy of using AI to extend what skilled staff can accomplish — not to reduce the level of skill required.
Why Foundational Knowledge Must Come First
The Difference Between Empowerment and Erosion
Dr. Amirfar draws a clear line between AI that empowers and AI that erodes competence. That line runs directly through the question of foundational knowledge. If a clinician or administrator knows how to do a task and understands it well, an AI tool makes them faster and more productive. If they do not have that foundation, the tool produces outputs they cannot evaluate, verify or stand behind. Furthermore, the inability to defend an AI-generated output is particularly dangerous in clinical and administrative settings where accountability is non-negotiable.
The Risk of Deploying AI Without Understanding
Simply using a large language model — including tools like ChatGPT — without the underlying knowledge to assess its output creates significant risk. Dr. Amirfar is direct on this point. If someone is not comfortable with a concept and does not understand how to perform the underlying task, asking an AI to do it does not solve the problem. Instead, it creates a new one: an output the user cannot critically evaluate. Moreover, in healthcare, outputs that cannot be discussed or defended intelligently have no place in clinical or administrative decision-making. The tool becomes a liability rather than an asset.
Healthcare’s Golden Period With AI
Dr. Amirfar describes the current moment as a golden period for healthcare AI. In this window, AI supports clinicians and makes them sharper without threatening their role. The technology handles the routine, the time-consuming and the pattern-recognition tasks — leaving humans to apply judgment, context and accountability. However, he acknowledges this period has a horizon. As AI capabilities advance, the question of whether tools can replace rather than support human roles becomes more pressing. That transition point, whenever it arrives, will require a fundamentally different set of conversations about AI’s place in healthcare.
Who Should Shape the Future Role of AI
Dr. Amirfar is clear that the direction AI takes in healthcare cannot be dictated from above. Healthcare professionals and patients must actively participate in shaping the role AI plays in care delivery. Decisions about AI’s scope cannot be imposed by government mandate or by insurance company policy alone. The people who deliver care and the people who receive it must have a voice in how AI integrates into their interactions. This call for inclusive governance reflects a broader concern that AI deployment driven by cost efficiency or regulatory pressure — without clinical and patient input — risks producing systems that optimize for the wrong outcomes.
