By K Yatish Rajawat,News18
Copyright news18
I travel to Mumbai frequently for work, and with my phobia of hotel rooms, I am always on the lookout for a place to stay. It takes me back to the time when I first came to Mumbai and was hopping from PGs to hostels while trying to become a journalist. The thing is that when I stay with friends or at clubs, I meet many people whom I would never have met otherwise.
Recently, I have noticed a new kind of uncertainty and stress among working professionals over 45 years old who hold mid- to high-level positions. If they are in the IT sector, they are concerned about being replaced by an AI-based system that has rendered most of their managerial tasks redundant. And when even storied IT companies, which had a history of never firing people, adopt American principles of AI-driven layoffs, the stress level is shooting up in a segment of society.
Mhatre is one such IT professional; he is four years away from retirement at an IT company, having worked his entire life. An IT company that never laid off employees and consistently maintained a large bench of talent. He confided that he is scared, as his peers have been asked to go on short notice. His son is studying at a US college; he faces high expenses and dreads going to the office, as he feels he is the next target for the HR axe.
The underlying cause of Mhatre’s stress is not that he is no longer useful to his company, but the company is taking a system view of everything in terms of output and input costs. The automation of large projects is so advanced that a system he ironically helped build is now allowing even a young manager to manage prestigious large projects. His expertise was valued, rewarded, and appreciated. Now, it has been commoditised into an AI platform that is changing the way his company operates and, more importantly, how CXOs think about middle management. I have often referred to this layer as email managers; it demeans their human contribution and also reduces them to a non-entity. This is the human cost of an AI platform, which is slowing down the country’s largest employment engine.
In this environment, it is essential to read the Niti Aayog’s report on AI. The focus of the report is to achieve a GDP growth rate of 8 per cent. This focus on just GDP growth for a subject like AI is myopic, especially for a government think tank that prides itself on independence and holistic development. It seems to be justifying AI in the guise of GDP growth but not factoring in the loss of jobs and the resulting fall in GDP.
The report says that trillions of dollars can be added by the faster adoption of AI. It also aims to facilitate the more rapid adoption of AI across sectors, resulting in productivity and efficiency gains. Blindly ignoring that these productivity gains will come at the cost of job losses, as few people do the job of many, and this may increase margins of companies, but will reduce earnings as the earning power of the population goes down. Whether this loss of employment and earnings is factored into the GDP gains is unclear, as the report does not address this issue.
NITI Aayog’s report, AI for Viksit Bharat, boldly acknowledges McKinsey & Company as a partner and cites multiple McKinsey/McKinsey Global Institute analyses that were used to build the report’s economic projections and scenario framing around seven pillars.
Several so-called potential outcomes have been enumerated in the India report (data platforms/AI Kosh, sovereign compute/GPU targets, skilling programs, regulatory sandboxes & model certification, ethics/safety/standards, sectoral deployments and PPPs/funding). These are approaches in the primary AI strategy of every country, including the US, the EU, and the UK as well. The difference is that in those countries McKinsey’s role is mainly limited to an external analyst whose published research is cited or used as background, not a credited partner of the official government strategy. This is especially alarming, as the role and impact of MNC consulting on government policy are significant. The conflict of interest that foreign consulting firms have with chip manufacturing and AI companies. Whether this is the reason that the report is silent on job losses or social impact on the country, we don’t know.
Analysing each potential outcome identified by the report, it also gives some links to the matching policy in the US / EU / UK documents for readers who want to go deeper into the comparison with other countries. The comparison is critical because India is not a developed country nor a small country. So, should our policies reflect and address our challenges and priorities as a young populous country, or should they blindly copy that of the US, UK, and EU?
Public datasets / data-as-a-strategic-asset (AI Kosh / national omics, Manufacturing Data Grid)
India: “Potential Outcome 1: India becomes the data capital of the world” — recommends anonymised, consent-based public datasets and sectoral data platforms (AI Kosh / national omics dataset / Manufacturing Data Grid). This is also part of the AI mission.
This is precisely what the EU-US are doing: The EU’s AI policy and guidance emphasize trustworthy datasets and data governance for AI (see the EU AI Act and Commission guidance on data governance). OSTP and White House materials also promote public-interest datasets and stewardship. See EU AI Act commentary and the White House EO and supporting materials. This link gives a quick summary of the EU report (softwareimprovementgroup.com). The EU passed the AI Act in May 2024, which applies not only to the risks and controls of AI within the EU but also globally. And this is the big difference: while the ambition and scope of the report is limited to creating data sets in India, the EU is trying to shape the global policy landscape.
The report overlooks the significance of sovereign data and its contribution to the global AI race. It does not address the data elephant in the room, which is how India should keep its data and the privacy of its citizens secured from the global AI platform. How should it regulate, control, and use data as a strategic resource as a competitive moat or an edge in an AI world? This myopia is conscious or unconscious; we don’t know, but it is the most essential aspect of any AI strategy so much so that companies have a policy for it, and there needs to be clarity on this issue for the nation. If this potential outcome of data capitalisation is expected to occur naturally, or if we open these APIs to the AIs of the world to scrape and steal, we don’t know, as the report remains silent on this front.
If it’s a strategic silence, not to show one’s cards, or if it’s just consultant silence, not to broach a subject that will affect its largest clients, the AI companies remain unknown.
Sovereign compute/GPU capacity
The second potential outcome is to “Establish a federated public infrastructure of high-end GPU clusters… mission targets deployment of 38k+ GPUs.”
US/EU parallels: US AI-related executive actions and the 2025 US Executive order on AI infrastructure emphasise national compute investment and resilience; EU industrial strategy materials discuss HPC and compute readiness. (See White House EO materials on compute and the Jan 2025 order on AI infrastructure). (The White House). Now the recommendation on the compute infrastructure is part of the global AI hype where data centres and investment are being fuelled by American chip and AI companies, which are ‘partnering’ with every country in the world to build their compute infrastructure.
Here, it is crucial to understand who will make this investment in GPU capacity. What will be the role of the private sector in this investment? Should India rush in blindly to build the same infrastructure as the US/EU, or should it adopt a different approach, like China’s? Which is building both the infrastructure and the AI models in a different way. China has recently refused even to buy Nvidia chips for its AI infrastructure. One strategic imperative here is not to replicate the Western AI models or infrastructure, or compete against them, as it is not just competition for GPUs and energy but also crucially for water resources. The tactical twist is to use their retail LLMs and compute. The strategic and the tactical are far more critical here.
Third Potential outcome: Skilling / lifelong learning
India: “Future skills — Train 50L+ students and professionals” and proposals for an AI Open University/AI Chairs. The whole debate on skilling is such a myth. Who are the people who will get skilled, what will be their skills, and what will they earn from these skills? The whole concept of prompt engineering skilling is surely being embedded into new AI models and has died a quiet death, even before it gained popularity. Any, even if it is a skill, it will remain a temporary one in the AI world.
Will Mhatre, at 45 years of age with experience in coding and management, be reskilled on AI and be able to earn the same? He will never be able to recover the same income level if he is fired from his current job.
There is another aspect of this skilling debate that needs to be clearly understood. If a highly skilled person, such as a PhD in Economics or genetics, is using the AI models, the more they use the AI engine, the faster the AI learns their mental models. As it learns the cognitive models, it has access to more data than a human can ever store in their head. It applies the same mental models and makes the skilled PhD person redundant. This is referred to as the commoditisation of skill levels, and, of course, this is nowhere mentioned in the report.
Of course, the report repeats what the UK/EU/US has said in their respective reports on skilling. The UK white paper and EU coordinated plans emphasise workforce reskilling and regulator capacity building; the US OSTP and other agency actions include workforce / reskilling measures. Whether any effort was made to consider the impact on existing skilled individuals, and whether the wages of highly experienced R&D engineers who will be replaced by younger, less experienced engineers will harm their careers.
Sectoral focus (health, finance, manufacturing, mobility)
The report focuses on banking, manufacturing, pharmaceuticals, and automotive (SAVs) with explicit enablers and numeric targets. The numerical targets are focused on efficiency gains that will result from productivity improvements directly related to the adoption of AI or in enhancing R&D.
While the reports in the US/EU/China all feature similar sectoral strategies in the exact domains, the EU’s rules, US agency guidance, and China’s AI planning documents prioritise identical sectors. (Examples: White House EO and EU AI Act sectoral guidance). (Federal Register). The problem is not the duplication of the strategy or the sectors; the challenge is the lack of foresight about these sectors and their role in job generation in India. In Europe and the US, most of these sectors have already outsourced their manufacturing to China.
These companies in their home markets only focus on design, engineering, and R&D. Indian companies or Indian promoters have never really adopted the habit of investing in R&D as they have always seen research as an expense item, which is the first to be cut if revenues slow down even marginally. While the report emphasises the pharmaceutical sector’s gains from applying AI to R&D, the companies that actually invest heavily in this sector are also limited. The impact of this advice is restricted.
The report includes the usual platitudes about startup funding and PPP models. India does not have the capital funding structure to invest the billions needed to create companies like OpenAI or Perplexity. The lack of specific models that should be funded may come from a paid report from a consulting firm.
K Yatish Rajawat is a public policy researcher and works at the Gurgaon-based think and do tank (www.cipp.in) Centre for Innovation in Public Policy (CIPP), and can be contacted on his https://www.linkedin.com/in/yatishrajawat/. Views expressed in the above piece are personal and solely those of the author. They do not necessarily reflect News18’s views.
