Deploy AI in healthcare with caution; build guardrails first, says Manipal CEO
While there are valid concerns around data protection in the use of artificial intelligence (AI) in healthcare, we must be careful in how we deploy AI, said Dilip Jose, CEO, Manipal Hospitals, on Thursday.
Speaking at ET's AI Conclave and Awards 2025, he said that healthcare is emotionally sensitive, morally complex, and context-heavy, making it critical for hospital chains to be mindful about where AI is embedded.
“In healthcare, we can’t be 99% accurate. The 1% error could mean a lost life. We can’t afford hallucinations in clinical settings,” he said. AI should be integrated into clinical workflows as a force multiplier rather than deployed as a standalone efficiency tool, he added.
In the same panel, Laina Emmanuel, cofounder of BrainSight AI, said AI is opening up possibilities to solve longstanding clinical problems. “There are so many unsolved clinical problems which, for the lack of computational infrastructure, were not addressed earlier. Today, with cloud infrastructure, we can run complex brain-mapping workflows and make them accessible to doctors and hospitals across the country,” she said.
She further added that informed consent must remain central to healthcare AI. “We had to take informed consent from every patient and ensure they understood how their data would be used. That is something other industries can learn from,” she said. Their team follows a risk-based approach and builds guardrails before deploying any solution in hospitals, she further added.
Abhijeet Vijayvergiya, cofounder of Nektar.ai, said enterprise AI deployments often fail due to poor data quality. Citing a study by the Massachusetts Institute of Technology, he said a majority of enterprise AI projects fail because of weak data foundations. “It’s not about which model you use. If the input data is poor, it’s garbage in, garbage out. Precision builds trust, and without trust there is no adoption or return on investments,” he said.
Hospital chains including Aster DM Healthcare, Yashoda Hospitals, Apollo Hospitals, and Global Health have recently partnered with or are in talks with AI startups for diagnostics, deploying AI-powered tools alongside human expertise to reduce turnaround time and improve accuracy.
Jose said the partnerships between hospitals and technology firms would be key to building validated models suited to India’s diverse population. “We have powerful longitudinal patient data across specialities and family histories. The dilemma is how to use it responsibly, balancing genuine patient interest with our role as a commercial organisation,” he said.
Speaking at ET's AI Conclave and Awards 2025, he said that healthcare is emotionally sensitive, morally complex, and context-heavy, making it critical for hospital chains to be mindful about where AI is embedded.
“In healthcare, we can’t be 99% accurate. The 1% error could mean a lost life. We can’t afford hallucinations in clinical settings,” he said. AI should be integrated into clinical workflows as a force multiplier rather than deployed as a standalone efficiency tool, he added.
In the same panel, Laina Emmanuel, cofounder of BrainSight AI, said AI is opening up possibilities to solve longstanding clinical problems. “There are so many unsolved clinical problems which, for the lack of computational infrastructure, were not addressed earlier. Today, with cloud infrastructure, we can run complex brain-mapping workflows and make them accessible to doctors and hospitals across the country,” she said.
She further added that informed consent must remain central to healthcare AI. “We had to take informed consent from every patient and ensure they understood how their data would be used. That is something other industries can learn from,” she said. Their team follows a risk-based approach and builds guardrails before deploying any solution in hospitals, she further added.
Abhijeet Vijayvergiya, cofounder of Nektar.ai, said enterprise AI deployments often fail due to poor data quality. Citing a study by the Massachusetts Institute of Technology, he said a majority of enterprise AI projects fail because of weak data foundations. “It’s not about which model you use. If the input data is poor, it’s garbage in, garbage out. Precision builds trust, and without trust there is no adoption or return on investments,” he said.
Hospital chains including Aster DM Healthcare, Yashoda Hospitals, Apollo Hospitals, and Global Health have recently partnered with or are in talks with AI startups for diagnostics, deploying AI-powered tools alongside human expertise to reduce turnaround time and improve accuracy.
Jose said the partnerships between hospitals and technology firms would be key to building validated models suited to India’s diverse population. “We have powerful longitudinal patient data across specialities and family histories. The dilemma is how to use it responsibly, balancing genuine patient interest with our role as a commercial organisation,” he said.
Next Story