(Editor’s note: This is the first part of a two-part series. The second part will be published in the March issue of Animal Health News and Views.)
Dave Barry, Pulitzer Prize winner for his commentary and consistent use of humor to present fresh insights into serious concerns, commented on what AI means for mankind. He declared, “AI is a computer thing that lay persons cannot possibly understand.” He went on to say, “AI enables college students to produce grammatically correct essays about books they haven’t personally read.” It is easy to understand the many questions and concerns veterinarians, physicians, and the public have about AI and their reluctance to adopt blindly.
Yet, AI has the capacity to revolutionize healthcare. AI can retain and review enormous datasets. It can identify patterns that elude humans. It possesses the potential to surpass human capabilities. Dr. Eric J. Topol, well-known cardiologist, scientist, and author, states, “The greatest opportunity offered by AI is the opportunity to restore the precious and time-honored connection and trust—the human touch—between patients and doctors.”
Physicians are noticing that AI grants them the opportunity to be fully present with their patients and focus more on the human aspects of the profession, allowing them to do more of what they love to do. In their offices, ambient AI can “listen” to interactions between them and their patients and automatically generate clinical notes for subsequent human review. Even before patients arrive physically at the doctor’s office, AI can enable intelligent triage that helps patients assess their symptoms and ensure they seek an appropriate path, all while relieving anxiety. AI-driven virtual assistants can “greet” clients at the virtual or physical front door and guide them through their clinical journey. The resulting conversations are more understandable by patients as medical jargon is avoided.
Today, we all use AI in our daily lives. AI is actively working when we order groceries online, get a ride, manage finances, and place an Amazon order; in fact, Amazon may have more immediate access to information about our packages than we have about our own health. What about the adoption rate of AI in healthcare by physicians and patients?
According to the American Medical Association in 2024, physicians are generally enthusiastic about the potential of AI in healthcare, with 68% seeing some advantage to use of AI in their practice and 66% reporting they use some type of AI in practice (up from 65% and 38%, respectively, in 2023). A 2022 study by Yale School of Medicine found that 55.4% of patients believed AI would make healthcare better and 66% deemed AI very important in their diagnosis and treatment. Key patient concerns are clinical accuracy and safety, privacy and security, human connection, equity, and trust that the system would use AI responsibly. A 2025 survey showed that physicians are adopting AI at a rapid pace and both physicians and patients see enormous promise in AI, but also potential limits.
We have not yet reached the point of being able to rely heavily on AI for diagnostics, considered the higher risk application of AI, although progress is being made. Research reported in 2026 shows that AI models often meet or exceed the diagnostic accuracy of physicians. Most progress has occurred in diagnostic imaging, for which AI often, but not always, outperforms human experts. A board certified radiologist is estimated to read 50-60 radiographs/day or 250/year. Typically, a radiologist is estimated to spend >1 to 5 minutes to interpret a standard radiograph and dictate a report. AI can read 3.9 million/year with each radiograph being interpreted within seconds.
Advances are being made in AI-assisted clinical decisions in other areas, as well. Here are a few examples. An article published in Nature in 2024 showed that computational pathology by Paige.ai achieved near perfect accuracy in diagnosing histopathology images. The purpose was to enable clinical decision support systems and precision medicine. Deep learning algorithms are being developed for dermatology. Google AI has a photo-based tool for smart phones for 288 conditions and reports that this tool equals the performance of specialists and outperforms generalists. For ophthalmology, IDx-DR, now LumineticsCore, uses AI in the diagnosis of diabetic retinopathy. A whole-body 3D imaging system images the entire skin surface (Vectra WB360), scanning and mapping abnormalities. AI is using massive data for predictive diagnosis that can outperform doctors and lead to earlier treatment.
AI seems to perform better with complex case analysis; for example, a 2025 study in the New England Journal of Medicine, reported that Microsoft’s AI Diagnostic Orchestrator correctly diagnosed 85% of complex cases, compared to 20% by a physician panel, which was >4 times the rate of the physician panel. This disparity is likely influenced by AI’s ability to assess large, complex data sets. In Shanghai, a showdown was held between four chief physicians from top hospitals and multimodal AI systems trained on thousands of medical cases. Diagnoses were made by physicians in 13 minutes and by AI in 2 seconds; however, there were some issues of diagnostic accuracy.
The results of AI diagnosis are very promising in controlled studies, yet to date not all have translated to real-world environments. One example is for stroke patients. Controlled studies demonstrated superb results for AI in imaging thrombosis. The results led some to ponder whether one would be liable for not using AI in a condition so reliant on speed and accuracy of diagnosis.
However, in practice 50% of AI diagnosed stroke cases were wrong. One potential explanation is that conditions in studies are tightly controlled, while those in the “real world” are not. Even standardization of images in practice is evasive. This example also raises a societal question about context of results. An accuracy rate of 85-90% may not be acceptable in developed countries, but may be excellent in developing countries in which no radiologists exist. This theory can be translated to Veterinary medicine, as well, especially in terms of rural communities and Veterinary deserts.
Clearly, AI adoption in human healthcare is accelerating but far from being flawless. AI must be used to augment, not replace human involvement and responsibility. This is one reason the term augmented intelligence is a more appropriate term than artificial intelligence.