Feature Article
By Nelumdini Samaranayake, PhD, Assistant Professor
Department of Medical Education and Health Systems Science at Texas College of Osteopathic Medicine, UNT Health Fort Worth
Sponsored by Robert Bunata, MD
This article was originally published in the July/August 2026 issue of Tarrant County Physician.
Artificial Intelligence (AI) is no longer something discussed only as the future of medicine; it is already part of everyday healthcare practice. From documentation and ambient scribing to decision-support tools, AI is increasingly integrated into the workflows physicians use each day. In many ways, these technologies are improving efficiency, reducing administrative burden, and helping clinicians manage growing amounts of clinical information. This article reflects the rapidly changing developments and evidence available as of May 2026.
As this technology continues to evolve in healthcare, an important question emerges: how will AI—and how should AI—influence the way clinicians think and make decisions? AI excels at identifying problems, processing vast amounts of data, detecting differences that might otherwise go unnoticed, and offering insights faster than any human could.
Cardiologist and author Eric Topol notes in Deep Medicine that, as technology becomes more capable of handling technical tasks, it may create more space for human-centered care in healthcare.1 While AI has the potential to support broader clinical insight and decision-making, human interpretation and communication, along with the ability to understand a patient’s real-life circumstances, cannot be fully replaced by technology.
What AI Does Not Fully Capture
What AI often struggles to capture is the bigger picture surrounding the patient.
Although newer AI systems are increasingly capable of analyzing conversational, behavioral, and contextual information, the challenge has shifted from simply processing data to meaningfully understanding the broader context of an individual’s life and circumstances. AI may detect patterns in speech, documentation, and other data sources, but it does not experience a patient’s circumstances firsthand. Family concerns, cultural influences, financial realities, and unspoken worries often require human interpretation and meaningful conversation to fully understand. These factors can be difficult to measure, and even when represented in data, they may not always be recognized within the context of an individual patient’s life.
As AI becomes more integrated into care, there is a growing risk of overly narrow clinical thinking, where complex human conditions are interpreted primarily through structured data. A patient can gradually become a set of variables rather than a person with a story. While this may improve efficiency, it can narrow clinical reasoning in ways that are not immediately obvious.
When Data Does Not Tell the Whole Story
This becomes particularly relevant in educational discussions when considering a patient with poorly controlled diabetes. An AI tool may appropriately highlight lab values and suggest adjustments to the treatment plan. From a data standpoint, that recommendation makes sense. However, even when social, cultural, and environmental factors are available, understanding how they influence a patient’s daily life can be more complex. Access to food, cultural dietary practices, health literacy, transportation barriers, financial constraints, and social support may affect whether a recommendation is realistic, acceptable, or sustainable for an individual patient. While AI may increasingly incorporate such information into its analyses, determining how these factors shape a patient’s choices and circumstances often requires human judgment, conversation, and shared decision-making. Without that broader understanding, even the most evidence-informed recommendation may not work effectively in practice. This is where thoughtful integration becomes important. AI should be viewed as a tool. Although powerful, it remains a tool. It is not the decision-maker, and it does not carry responsibility or accountability. Those responsibilities remain with clinicians. The concern is not the presence of AI itself, but how increasing reliance on these tools may skew clinical reasoning over time.
Trust in AI is not only a technical challenge but also a legal one. Recent healthcare AI discussions published in JAMA have highlighted that regulatory and accountability frameworks are still evolving, leaving important questions unanswered regarding accountability, oversight, and liability when AI-supported tools contribute to errors or patient harm. As AI becomes increasingly integrated into healthcare decision-making, establishing clear standards for accountability will be essential for maintaining public trust.2
The Responsibility Still Belongs to Us
We are already seeing AI embedded in documentation workflows, particularly through the use of ambient scribing technologies. These tools can generate clinical notes, summarize encounters, and even suggest language for patient communication. They may reduce administrative burden and allow physicians to spend more time interacting directly with patients, which is a meaningful advancement. However, their use also requires careful clinical judgment and oversight. There have been situations in which AI-generated outputs were shared with patients without being fully reviewed or validated. In some cases, information that appeared to represent a diagnosis or clinical conclusion created confusion or unnecessary distress when it had not been carefully verified by the responsible physician. These moments are not failures of technology alone; they are reminders of how easily shortcuts can influence communication and clinical interpretation. They reinforce an important point: anything generated by AI must be reviewed, interpreted, and confirmed before it is shared with a patient. This includes not only the accuracy of the information itself, but whether it is appropriate for the situation and whether the patient has been adequately prepared.
Clear communication and informed patient consent are not optional steps; they are essential components of care. AI does not remove that responsibility; it reinforces it.
Why Clinical Judgment Matters More
As healthcare environments become increasingly influenced by AI, the role of the clinician becomes more important. AI should not come first in the clinical process. Instead, it should come after an initial, patient-centered assessment. First, we engage with the patient; we listen, observe, and begin to understand their story. Then we form an initial impression grounded in both data and human interaction. Afterward, AI may be used to expand or challenge that thinking. This sequence matters because it preserves clinical reasoning while allowing AI to serve as a cognitive support tool rather than a directive force. This approach helps keep clinical decision-making focused on the patient rather than solely on algorithmic output. Equally important is maintaining a mindset of critical evaluation.
AI systems are not perfect. They can generate highly confident responses that are incorrect, particularly when built on biased, incomplete, and non-representative data. They may produce outputs that appear highly credible but lack important context or accuracy. In those moments, the clinician must be willing to pause, question, and reassess. This is not about distrust, but responsibility. The presence of AI requires us to think more carefully, not less. It challenges us to ask better questions: Does this output align with what I am seeing? What might be missing? What does the patient’s situation reveal that the data does not? These are the questions that protect against over-reliance and preserve sound clinical judgment.
Looking Ahead
The role of AI in medicine goes beyond technology. It influences how clinicians think, how decisions are made, how information is presented, and how care is delivered. As AI continues to grow within healthcare, it brings both valuable opportunities and new responsibilities.
The benefits are promising. AI can assist clinicians in processing information, recognizing patterns within datasets, and highlighting details that may otherwise go unnoticed. When used thoughtfully, these tools can strengthen clinical assessment and support decision-making without replacing professional judgment. But the risk is less visible. If we are not careful, AI can narrow our thinking. Clinical attention can gradually shift toward structured data while real-life circumstances receive less consideration. Over time, this can influence how patients are understood and how decisions are made.
This issue is not rooted solely in AI. In many ways, the outcome depends on how thoughtfully these tools are incorporated into clinical care. It also creates an opportunity to shape how future clinicians think about AI—not as a shortcut or an answer engine, but as a tool that requires interpretation, reflection, and responsibility. We can guide this use of AI to strengthen clinical reasoning rather than weaken it. AI may assist with information analysis, but medicine continues to rely on human interpretation, empathy, and trust. Medicine is not defined by competing with machines in processing data. That is not where our value lies. It is in the human qualities that machines cannot replace. We understand nuance. We interpret context. We recognize emotion. We build trust. We see the patient not as a set of data points, but as a person with a life, a story, and circumstances that matter.
AI may assist in clinical decision-making, but accountability, interpretation, and patient care remain human responsibilities. While AI can analyze data, clinicians are still responsible for seeing the whole person. AI is already a part of everyday medicine; now the question is how AI will shape the way we think, decide, and care.
References
- Eric Topol, Deep Medicine: How Artificial Intelligence Can Make Healthcare Human Again (New York: Basic Books, 2019).
- Michelle M. Mello and I. Glenn Cohen, “Regulation of Health and Health Care Artificial Intelligence,” JAMA 333, no. 20 (2025): 1769–1770, https://jamanetwork.com/journals/jama/fullarticle/2831831.




