AI in healthcare sees slow but steady rise: Why top doctors say AI can’t replace clinical intuition

“The main challenges include the high cost of integrating Generative AI models, the lack of governance, and the protection of sensitive patient information. There is also a lack of a standardized Clinical Decision Support System (CDSS) framework. Medical diagnosis often relies on combining multiple clinical, laboratory, and radiological clues. For instance, ECG changes, raised blood markers, chest pain, and elevated heart rate may together indicate a heart attack. As a doctor with clinical experience, I would be able to put these pieces of a puzzle together. However, translating this complex reasoning into AI algorithms can be difficult, inaccurate, and not representative of the varied population in general. Some of these algorithms lack explanation ability, and doctors question whether such systems can ever provide 100% certainty,” Dr Shukla observed.

According to Dr Shukla, another limitation is data. “For AI models to produce consistent and reliable outputs, they must be trained on millions of standardised data points. However, hospital data is often not shared, and there is a lack of uniform data formats. This restricts effective training and scalability. There is also the issue of the “black box” nature of AI. Doctors need to understand how a model arrives at a conclusion,” he said. In serious conditions like cancer or stroke, an AI-generated output without clear reasoning may not be trustworthy. This lack of transparency limits the adoption of AI for critical decision-making, Dr Shukla explained.

AI is most useful in simpler, repetitive tasks

Presently, several doctors note that AI is most useful in simpler, repetitive tasks. Dr Vijay Natarajan, cardiac surgeon at Poona Hospital, said that the use of AI in healthcare is currently limited but steadily evolving. “AI is being explored for tasks such as generating discharge summaries for long-stay patients and assisting with ICD coding, where each diagnosis is assigned an international classification code for every patient. It is also being considered for preparing ICU transfer-out summaries, helping streamline documentation and improve efficiency,” the cardiac surgeon said.

Dr Prasad Rajhans, chief intensivist at Deenanath Mangeshkar Hospital, too emphasised the importance of backend software in healthcare, particularly for managing discharge summaries, patient records, and overall clinical documentation.

“AI integration into these systems can significantly reduce administrative workload and improve efficiency in routine medical tasks,” he said.

According to Dr Meeta Nakhare, consulting gynaecologist,  AI in healthcare can be beneficial in cases of limited resources like manpower. “For instance, in comparison to a faulty blood pressure apparatus, an automated one, which recalculates the patient’s pulse, BP, and instantly gives a report, can help save on resources, time, money, and energy. But we have to remember that there has to be standardization of AI for use in healthcare. We don’t want a fake model that gives us wrong reports,” Dr Nakhare said.

‘AI in diagnostics is promising in diabetic retinopathy’

Dr Aditya Kelkar, Director, National Institute of Ophthalmology (NIO), also agreed that AI is used in hospital management for appointment scheduling, allocating patients to various sub-specialty consultants. However, he acknowledged that AI in diagnostics, especially for age-related macular degeneration and diabetic retinopathy, is promising.

“AI is being used in cataract and retina surgeries mainly to effectively control eye pressure during the procedure to minimise complications,” Dr Kelkar said.

Ankit Modi, founding member and Chief Strategy and Growth Officer at Qure.ai, a healthcare firm, further pointed out that AI is becoming an integral part of everyday diagnostics, not as a replacement for clinicians but as a powerful support system. According to Modi, AI enables faster interpretation of scans and can be particularly impactful in settings where access to specialist expertise remains limited.

Trust, empathy, communication heart of medical practice

Dr K M Paknikar, ANRF Prime Minister Professor at COEP Technological University, said that clinical decision-making is not merely a computational exercise. “It is a nuanced synthesis of scientific knowledge, experience, intuition, ethical responsibility, and an understanding of the patient as a whole person rather than a dataset.

“AI can assist by analysing vast volumes of data, suggesting diagnoses or identifying patterns that may escape the human eye. But it cannot assume accountability nor can it fully grasp context, uncertainty, or emotional and social dimensions of illness,” Dr Paknikar said.

“A physician must weigh not only what can be done but what should be done, often in situations where evidence is incomplete or lacking. Trust, empathy, and communication, which lie at the heart of medical practice, cannot be outsourced to algorithms. In this sense, AI will function as a powerful instrument, but the clinician will remain the final arbiter,” he added.