Ildiko Csiki
A Defining Moment for AI in Healthcare
TIME Magazine announced its 2025 AI Architects today, a group of individuals selected for their transformative influence on artificial intelligence and the National Academy of Medicine recently published a commentary that deserves equal attention. The piece, titled "Critical AI Health Literacy as Liberation Technology," articulates with unusual clarity the fundamental tension that will define healthcare's technological future.
The question is not whether AI will reshape medicine. That transformation is already underway. The question is who that transformation will serve.
Two Divergent Paths
The National Academy of Medicine commentary draws a line that healthcare institutions, policymakers, and patients must acknowledge. On one side stands institutional AI: systems designed to standardize clinical workflows, enforce payer compliance, manage organizational risk, and extract value from patient data. On the other stands patient-directed AI: technology that shifts control toward the individual, offering opportunities for personal autonomy not constrained by institutional policy.
This distinction matters because most AI deployed in healthcare today serves institutional priorities. Algorithms flag insurance claims for denial before a human reviews them. Triage models, sometimes built on training data that reflects historical biases, determine which patients receive timely attention and which wait. Prior authorization systems create barriers between patients and treatments their physicians have recommended. These systems operate largely without transparency. Patients experience their effects without understanding their mechanisms.
The authors of the NAM commentary introduce a concept worth considering carefully: the emergence of "AI Patients." These are individuals who use technology not merely to receive information but to sharpen their reasoning, deepen critical engagement with their care, and reclaim ownership of their health narrative. They are not passive consumers of algorithmic outputs. They are active participants who leverage AI as a tool for intellectual and medical self-determination.
The Institutional Limitation
Health systems face genuine constraints. Electronic health records must maintain data integrity. Clinical decision support must meet regulatory standards. Liability concerns shape what information can be communicated and how. These constraints are real, and dismissing them serves no one.
Yet these same constraints can produce systems that prioritize institutional protection over patient need. When an algorithm denies a claim, the patient bears the consequence while the institution bears reduced risk. When a triage model reflects the biases embedded in its training data, marginalized patients experience worse outcomes while the system maintains its operational metrics.
The gap between what institutions need and what patients need is not always wide. But when it exists, patients rarely have tools to see it clearly, let alone address it.
What Patient-Directed AI Makes Possible
Consider what becomes available when patients have access to AI that serves their interests directly. They can ask questions that institutional systems might not surface: What clinical trials exist for my specific situation? What do survival statistics actually mean for someone with my characteristics? What questions should I be asking that I have not thought to ask?
They can cross-reference recommendations against independent sources. They can explore treatment options that their particular health system may not offer. They can prepare for clinical encounters with greater knowledge and sharper questions. They can find emotional support at 2am when the clinical system is unavailable and the fear becomes overwhelming.
Patient-directed AI does not replace clinical expertise. It augments patient agency. It creates conditions where the patient-physician relationship can function as a genuine partnership rather than an information asymmetry.
The Oncology Imperative
For cancer patients and caregivers, these considerations are not theoretical. They are matters of survival. The ability to understand treatment options, identify relevant clinical trials, ask informed questions during appointments, and access emotional support throughout treatment can influence outcomes in ways that clinical systems alone cannot provide.
At Wondrlink Foundation, we hear from patients daily through our WondrVoices community. What strikes us consistently is their refusal to accept passivity. They want tools that respect their intelligence. They want information presented honestly, including the uncertainties. They want support that meets them in their actual circumstances rather than in some idealized version of what a patient experience should look like. Our #cancervoice campaign is designed to ensure everyone has a voice to shape the future of cancer care.
Another tool, WondrGPT is being built as conversational AI designed around patient and caregiver needs while integrating appropriately with institutional workflows. Our objective is to create agentic AI companions that provide psychosocial and emotional support for oncology patients and caregivers, with clinical evidence demonstrating effectiveness.
The emphasis on evidence matters. Patient-directed AI cannot succeed as a consumer technology disconnected from clinical reality. It must demonstrate measurable benefit. It must integrate with care rather than fragment it. It must maintain appropriate boundaries while still serving patient autonomy.
Building the Evidence Base
The NAM commentary identifies a critical gap: health systems currently lack the infrastructure and orientation to deploy AI that genuinely serves both institutional and patient interests. Closing this gap requires more than technological innovation. It requires new models of validation, new frameworks for integration, and new commitments from institutions willing to share some measure of control.
For patient-directed AI to achieve legitimacy within clinical settings, it must produce evidence that institutions can trust. Randomized trials. Measurable outcomes. Demonstrated safety. This is the work ahead.
Through initiatives like our forthcoming cancer voice campaign, we intend to ensure that patient perspectives shape the development of these tools from the beginning. Patient-directed AI built without patient direction would be a contradiction in terms.
The Architects of This Moment
TIME's selection of AI Architects for 2025 reflects a recognition that we are living through a period of technological consequence. The individuals shaping AI systems today are making decisions whose effects will persist for decades.
In healthcare, the architectural choices matter profoundly. Systems designed primarily around institutional efficiency will produce one kind of medicine. Systems that balance institutional needs with patient agency will produce another. The technology is not neutral. The choices embedded in its design reflect values, and those values will shape millions of lives.
The National Academy of Medicine has articulated the opportunity. The question now is whether health systems, technologists, policymakers, and patient advocates will build the tools and the evidence base that make patient-directed AI a clinical reality.
At Wondrlink Foundation, we believe they must. Because every patient deserves not only access to care but meaningful participation in the decisions that shape it.
Now it's the time to build the tools and the evidence base that make patient-directed AI a clinical reality.
Topics: