By Satyendra Josh who is part of the Statistics and Data vertical at SETU Aayog, the Government of Uttarakhand’s policy think tank. His work focuses on strengthening data systems, supporting evidence-based policymaking, and contributing to the design and implementation of technology-enabled governance initiatives.
In February 2026, during the AI Summit, the Ministry of Health and Family Welfare released SAHI (Strategy for Artificial Intelligence in Healthcare for India). Amongst the five pillars mentioned in the document, in Pillar 1 (Governance, Regulation, and Trust) , while talking about Equity and Inclusivity it warns that failing to ensure representative training data can lead to “systematic under-prioritisation of certain groups, reinforcement of structural disparities, and the erosion of public trust.” In Pillar 2 (Health Data and Digital Infrastructure) it further states that “inconsistent documentation, missing fields, variable structures, and weak provenance undermine AI performance.” In other words, the strategy document explicitly states what can go wrong in deployment and use of AI.
A major implication of this structural gap is that AI systems in healthcare will not operate in a vacuum. Instead, they will be layered into service/data delivery structures that exist already within an organization. But records from Himalayan states frequently show gaps, missing entries, or inconsistencies in reporting key positions and facility level information. When basic staffing details are not systematically documented, it becomes difficult to build datasets that truly reflect conditions on the ground. This leads to incomplete digital records, gaps in patient care information, and very little data to properly deploy and improve AI systems.
The government’s own health statistics, Rural Health Statistics 2021–22 and the newly released Health Dynamics of India 2022–23, tell a different story from SAHI’s aspirational narrative. In many rural Primary Health Centres across Uttarakhand, essential health staff are simply not available in adequate numbers. Several facilities function without a full time doctor, and in many cases even nursing support is missing. As a result, frontline service delivery often depends on limited personnel managing multiple responsibilities at once.
In such contexts, AI risks becoming a superficial add on to an already fragile system rather than a real driver of change. Therefore, the success of AI tools for triage, decision-making, or prediction depends not just on how reliable the technology is, but also on whether trained staff are available to record data, understand it, and take timely action.
SAHI correctly states that AI can enable earlier detection of disease, more precise clinical support, improved targeting of public health interventions. Every one of those promises is true, but in a setting where a proper manpower exists to act on the output, where a device can be charged, and where data has been recorded. The rural and far flunging areas in the Himalayan states, by and large, do not yet meet those conditions.
In order to realise the picture that SAHI shows us we must first create a good system to make sure health data is correct and complete. But such a system is needed before using Artificial Intelligence, not along with it. Right now, this system is still not fully ready, and the data in many places is still not good enough.
Therefore, the order of work should be simple. First, make sure that Sub Centres have proper electricity so that basic services can run. Then repair and reopen Community Health Centres that are not working well. Deploy the manpower and referral transport for the remotest of areas. After that, build proper digital health records as promised under Ayushman Bharat Digital Mission.
Only then should Artificial Intelligence tools be used. In parts of the Himalayan regions where these basic facilities are already strong, testing Artificial Intelligence tools can begin now. But in places where there are still fewer doctors, fewer working health centres, and weak basic services, it is better to first improve hospitals, staff, transport, electricity, and record keeping. Artificial Intelligence can wait.
SAHI’s third pillar talks about building the skills of health workers so that they can understand and use Artificial Intelligence. This means adding basic knowledge about Artificial Intelligence into the training of frontline workers. This is a very practical and implementable idea for all Himalayan states. When new technology is introduced, the people who actually use it must feel confident and comfortable with it. But while planning such technology, it is important to think first about the field level officers (such as ANM, ASHA etc.) who works in villages and remote areas. Many technology tools are designed for doctors sitting in big hospitals, but the real need is often at the village level. These tools should be simple to use, available in local languages (like Garhwali and Kumaoni in Uttarakhand), and should work even where internet connection is weak or electricity is not always available.
The tools should be tested in the same difficult conditions in which frontline workers actually work. For example, they should work in small Sub Centres where power cuts are common, where devices cannot be charged every day, and where one worker may be handling many tasks at the same time. If a tool only works well in large district hospitals with good facilities, it may fail in the places where it is needed the most.
For many years, the government has worked hard to build a health workforce that reaches people in distant villages and continues to serve even in tough situations. A good technology plan, therefore, should carry forward this effort. It should create technology that truly supports these workers, makes their job easier, and helps them care for people better. At the same time before moving towards deployment of AI in a Himalayan ecosystem, we should ensure availability of proper infrastructure and manpower to sustain AI solutions and have meaningful impact.
