This article is part of our Ask the CMO series, where Cook Medical’s chief medical officer, Dr. John Kaufman, answers questions. Learn more about Dr. Kaufman in his Meet Our Leaders bio.
Key highlights:
- Artificial intelligence offers both advantages and challenges for the medical field.
- Some of the advantages include data synthesis, disease diagnosis and management and improved imaging.
- Some of the challenges of AI in the medical field include data privacy and security risks, potential algorithmic bias and questions about liability.
- However AI is integrated into healthcare, it will never change replace the need for human empathy.
It is hard to go to a medical meeting or open a journal without seeing something about artificial intelligence, which reflects how broadly AI is reshaping nearly every aspect of modern life. At Cook, we are being exceptionally proactive about understanding and responsibly incorporating AI into what we do, and I think that forward-looking posture puts us in a strong position as the technology continues to evolve.
In medicine broadly, AI will prove to be a powerful tool for supporting and aiding in diagnosis and clinical management. But humans will ultimately have to decide what to do with the information it surfaces. That distinction matters more than most people realize.
Ways AI can accelerate the medical field
Working through data
One of the clearest near-term opportunities for AI in medicine is its ability to analyze volumes and types of data that are genuinely difficult for the human mind to process efficiently. A physician seeing thirty patients a day cannot be expected to synthesize every relevant data point from a patient’s full history, recent lab trends, imaging findings, medication interactions, and emerging clinical literature all at once. AI can do exactly that and present the output in a way that is useful immediately.
The critical caveat is that AI is only as good as the data that it works with. If information is not available in a digital format, if it is conflicting across records, or if it is simply inaccurate, a human clinician still has to make the call as to whether the AI output is useful or not. The tool informs, but does not replace, human judgment.
Diagnosing and managing disease
AI is already demonstrating meaningful potential in disease diagnosis and management, particularly in areas where pattern recognition across large datasets is critical to accuracy. In oncology, AI models are being used to identify subtle disease markers in tissue samples and lab results that might otherwise be missed or caught at a later, less treatable stage. In cardiology, algorithms are being trained to predict adverse events before they occur by identifying patterns in patient vitals, ECG data, and clinical history.
The promise here is not that AI will replace the clinician but that it will give clinicians a more complete picture at the moment they need it most, improving both the speed and the accuracy of diagnosis.
Imaging
Medical imaging is perhaps the area where AI has already made the most visible inroads, and where the opportunity is most immediate. AI-powered imaging analysis tools are now capable of detecting abnormalities in radiology scans, pathology slides, and other diagnostic images with a level of consistency and speed that complements what radiologists and other specialists can do on their own. Studies have shown that AI-assisted image review can reduce the rate of missed findings and help prioritize the most urgent cases for faster human review.
At Cook, imaging plays a critical role in many of the procedures and therapies we support, and we see AI-enhanced imaging as an area where the opportunity to improve patient outcomes is particularly significant. One prominent example is interventional MRI (iMRI). Cook’s complete iMRI suite with Siemens Healthineers includes AI to improve the image quality of soft tissue and small structures. Additionally, the data from the iMRI suite will be compiled and used to learn about how to improve the efficacy of interventional procedures.
Challenges AI will bring
As genuinely exciting as these opportunities are, it would be a mistake to discuss AI in medicine without being honest about its limitations and the challenges it introduces.
Data privacy and security
AI systems in healthcare require access to enormous amounts of sensitive patient information to function effectively, and the responsibilities that come with handling that data are significant. The FDA released best practices specifically for the medical device industry on how to keep data safe. The best practices include data quality assurance, data management and cybersecurity considerations throughout the medical device product lifecycle.
Health systems and medical device companies alike will need to invest heavily in governance frameworks that ensure AI tools are deployed ethically and in compliance with evolving regulatory standards. The stakes here are high, and the margin for error is low.
Algorithmic bias
AI models are trained on historical data, and if that data reflects existing disparities in healthcare access or treatment, the model will replicate and potentially amplify those disparities. The PLOS Digital Health journal published a review that discussed the need for AI data to be checked for bias. The article gave examples such as an algorithm used in US hospitals that was biased against black patients in resource allocation. Also, dermatological AI showed lower diagnostic accuracy for conditions like melanoma in darker-skinned individuals due to the AI agent being trained primarily on fair-skinned images.
Ensuring that AI tools are trained on diverse, representative datasets is not just a technical requirement. It is an ethical one that the field needs to take seriously before these tools become more deeply embedded in clinical workflows. A recommendation generated by a biased model is not a neutral recommendation, and the consequences in a medical context can be serious.
Liability and accountability
When an AI system informs a clinical decision that leads to a poor outcome, the question of who bears responsibility is not yet fully resolved from a legal or regulatory standpoint. Is it the clinician who acted on the recommendation? The company that built the model? The health system that deployed it?
These are not hypothetical questions. They are conversations that the medical community, regulators, and companies like Cook need to be actively part of as the technology matures and its role in clinical decision-making expands.
The American Medical Association released a statement in 2025 about the need for physicians to be involved in the entire AI process. They also advocate for protections to make sure that liability for AI is appropriately apportioned and limits physician liability for AI errors and performance issues. As Cook Medical works with physicians and incorporates AI into our own medical device development process, we need to think about the potential consequences.
The Thing AI Will Never Be Able to Replace
Perhaps the most important limitation of AI in medicine is also the most fundamental. AI does not have true empathy. Although interactions with AI can feel very personal and are always framed in positive language, these are stylistically programmed rather than genuine responses. AI models do consider the actual person the same way that medical professionals do, or the way that we at Cook think about every patient whose care our products touch. AI can surface a recommendation, but it cannot hold a patient’s hand. It cannot read the room when a family is frightened. It cannot interpolate the clinical data with a patient’s values, preferences, and quality of life in the way that a skilled, compassionate provider can.