Elsevier's Jan Herzhoff says that critical thinking, not AI, will guard against de-skilling

Concerns about doctors losing skills due to overuse of artificial intelligence (AI) can be reduced if they combine AI tools with critical thinking, instead of blindly trusting AI responses, said Jan Herzhoff, president of global health businesses at Elsevier.

Speaking at the India AI Impact Summit 2026 (February 16–20), Herzhoff referred to a 2025 study published in The Lancet Gastroenterology and Hepatology. The study found that experienced healthcare professionals were 20% less likely to detect benign tumours during colonoscopies three months after they began relying on AI assistance. According to Herzhoff, de-skilling happens when clinicians depend only on short AI-generated answers without thinking deeply or expanding their own knowledge.

He explained that doctors are often very busy. When they have a question, they may quickly ask an AI tool, check the reference briefly, and then move to the next patient. If they stop applying critical thinking and do not build new connections in their understanding, their skills may gradually weaken.

The summit, inaugurated by Narendra Modi at Bharat Mandapam, focused on making AI more inclusive and accessible. The event saw participation from representatives of over 100 countries and more than 500 global AI leaders.

Herzhoff said AI is at a crucial stage where it can both benefit society and create challenges. From a regulatory and business perspective, AI has a major influence on how societies function, making it important to discuss its responsible use.

Elsevier is collaborating with the Indian government on a project called ‘Digital Innovations and Interventions for Sustainable HealthTech Action’ (DIISHA), aimed at digitally empowering ASHA workers. ASHA (Accredited Social Health Activist) workers are community healthcare workers under the National Health Mission launched in 2013.

Elsevier has developed an AI tool called ‘ClinicalPath Primary Care’ (CPPC) to help ASHA workers with clinical decision-making. The tool is designed to reduce the healthcare gap between urban and rural areas by bringing expert-level screening support to remote regions.

A pilot study in Uttarakhand’s Dehradun district involved 20 ASHA workers from a primary health centre in Raiwala. They were trained to use CPPC and monitored for 12 weeks. According to a report prepared with contributions from the department of community medicine at All India Institute of Medical Sciences Rishikesh, CPPC improved the confidence, efficiency and adherence to clinical guidelines among ASHA workers. However, the report recommended more training, inclusion of additional diseases, and better offline functionality to improve its usefulness.

Elsevier has also developed another AI tool called ‘ClinicalKey AI’, launched in November 2023. It is a conversational search tool designed to assist doctors in clinical decision-making. Herzhoff said around 300 hospitals globally are using ClinicalKey AI, and adoption increases significantly when it is integrated into electronic health record (EHR) systems. However, India’s fragmented EHR systems present challenges for wider adoption.

On the issue of medical misinformation, Herzhoff stressed the importance of teaching critical thinking to medical students early in their education. He also suggested building “trust markers” into AI systems — such as highlighting whether information comes from high-quality journals or showing impact factors — to help clinicians distinguish reliable sources from poor-quality content.

He concluded that AI can greatly support healthcare, but only if it is used responsibly, alongside human judgment and strong institutional support.