India is Scaling AI Skills. But the Jobs Data Is Disappearing

India’s skilling machine is getting bigger, faster, and more technologically ambitious. At the same time, it’s getting harder to track skilling outcomes.

A fresh Lok Sabha reply from the Ministry of Skill Development and Entrepreneurship (MSDE) shows that the government is expanding AI, machine learning, robotics, cloud computing, data analytics, cybersecurity, and other emerging technologies through the fourth phase of the Pradhan Mantri Kaushal Vikas Yojana (PMKVY). The government says it has also developed 170 National Skills Qualification Framework (NSQF)-compliant courses, including 32 future-skills courses, through the Craftsman Training Scheme.

Yet, what is a more troubling revelation, the Ministry also admitted that it was no longer monitoring placements under PMKVY 4.0. The government reasoned that since PMKVY 4.0, operational since FY2022-23, is focused on enabling trainees to choose different career paths, it is not tracking placements.

The result is a peculiar asymmetry. India is operating on the fast lane to certify students with new-age skills, while turning a blind eye to whether those skills translate into employment.

From millions of certificates to AI learners

The scale of the earlier PMKVY programme is substantial. Between 2015 and 2022, the first three phases had an outlay of around ₹14,450 crore, with 1.11 crore candidates certified from a utilisation of ₹9,261 crore. The government now says 24.38 lakh candidates were placed under the Short-Term Training (STT) component.

However, the CAG's audit presented a more complicated picture. It flagged that only 23.18 lakh candidates, or 41% of those trained under the STT component, were placed.

The audit also found that training did not consistently align with job-role-specific or state-level skill gaps. Placement verification itself was problematic. In Kerala, the CAG found incorrect placement documents submitted by training partners that led to the recovery of ₹22.33 lakh. 

The report also identified weak controls around beneficiary identities and bank details. Almost 94.53% of candidate records examined had blank bank account details, while thousands of bank accounts were associated with multiple beneficiaries. The audit also found apparently invalid account numbers and other unreliable entries. 

That history makes the decision not to monitor placements under PMKVY 4.0 particularly significant.

As Sumesh Nair, Co-founder of upskilling platform Board Infinity, puts it, AI-skilling programmes should be measured through a “verified employment-and-performance scorecard”, rather than course completions or certificates. He tells AIM that this should include employment or self-employment at six and 12 months, job retention, wage progression, employer satisfaction, industry projects or apprenticeships, and independent competency assessments.

“For AI roles especially, the meaningful test is whether graduates can apply skills in real workflows with limited remedial training, rather than whether they have completed a course,” he stresses. 

AI-Skilling Boom is Highly Uneven

The data submitted in the Lok Sabha offers a detailed picture of how PMKVY 4.0's AI and emerging-technology training is spreading across states.

Roughly 5.4 lakh candidates were trained in future-forward skills between FY2023-24 and June 2026. The programme recorded around 19,500 such trainees in 2023-24 before surging to more than 4.48 lakh in 2024-25. The figure then falls sharply to around 45,000 in 2025-26, with 2026-27 standing at roughly 31,000 by June 2026. 

The state-wise distribution has been dropping.

Uttar Pradesh reported more than 1.03 lakh AI/emerging-tech trainees across the period, including 88,868 in 2024-25. Rajasthan recorded 84,144, while Madhya Pradesh reported around 67,064. Tamil Nadu, Karnataka, Punjab, and Bihar follow at considerably lower levels.

The number of candidates trained dropped dramatically in the following year. Tamil Nadu trained 7,086 candidates in 2025–26, followed by Uttar Pradesh at 6,672, Madhya Pradesh at 5,260, Karnataka at 4,259, and Punjab at 3,356.

The government does not explain in the parliamentary answer why reported PMKVY 4.0 AI/emerging-tech training fell by roughly 90% in 2025-26. More importantly, there is no corresponding state-wise placement picture for these trainees.

Learning Itself is Changing

The industry is seeing a different evolution when it comes to not just only the kinds of skills that are in demand, but also how one obtains them.

Shronit Ladhani, Founder of AI-led learning platform LearnTube by CareerNinja, says learning is moving “from the degree to the course to the task.” Learners increasingly want AI training that solves an immediate workplace problem rather than a generic qualification.

“The fastest-growing course isn't ‘learn AI’. It's AI plus the job you already have. Marketers running campaigns with AI, analysts automating their reporting, HR teams screening with it,” he highlights.

Ladhani also believes that the “demand frontier has already moved past prompts to agentic AI.”

Apurv Mehra, Co-founder and CEO of edtech startup Disha AI, sees a similar shift. Generative AI fundamentals, prompt engineering, and hands-on familiarity with tools such as ChatGPT, Claude, Midjourney, and NotebookLM are seeing strong demand. Learners, however, are primarily interested in practical workplace applications.

That creates a challenge for a government system traditionally built around courses, certifications, and training numbers.

“Certifications are losing their signal value. Degrees are headed the same way,” Mehra says. Increasingly, what matters is proof of work—whether someone can build something, solve a real problem, and think from first principles.

“The market is already moving this way; employers are asking ‘show me what you've built’ far more than ‘show me what you've passed,” he adds.

AI Literacy is Not AI Employability

The distinction becomes even more important as India attempts to move millions of people into an AI-enabled economy.

Nair says companies can find candidates with basic exposure to AI tools, but struggle to find people who can work with data, assess AI outputs, understand business applications, follow responsible-AI and security practices, and deliver practical projects.

“AI literacy is not the same as AI employability,” he says. “Industry should be a co-owner of quality, not simply the recipient of trained candidates.”

The government has recognised the need for industry alignment, establishing institutions such as Sector Skill Councils to study skills gap, District Skill Committees to identify local employment opportunities, and the SANKALP programme, which supported a national skill-gap study by NCAER. It also points to partnerships with IBM, Cisco, Microsoft, and AWS.

But the CAG's findings show why the existence of skill-gap studies does not necessarily guarantee effective training. Earlier PMKVY phases suffered from weak micro-level skill-gap mapping and training that did not consistently correspond with available demand data. 

The issue, therefore, is less about the number of reports, partnerships or courses and more about whether those inputs improve the quality and durability of employment.

“The real skills story of this decade is getting that same rigour into the other 90%. That's the gap we're building for,” Ladhani adds.

The Risk of the Certificate Economy

Alisha Butala, Policy Analyst and Researcher, argues that India's deeper problem is not simply a shortage of AI-ready talent. It is that the education and skilling system is designed to respond to innovation rather than produce it.

“AI is the latest example,” she says. 

An AI-ready India cannot simply mean training people to use technology developed elsewhere. India is racing to make students AI-ready while the technology frontier is already moving towards quantum computing and other advanced technologies. 

That concern echoes Niranjanaradhya's criticism of the broader skilling architecture, as reported by AIM in November 2025. The programme head of universalisation of education at the National Law School of India, Bengaluru, argued that without a clear vision of the target group, the purpose of training, and the opportunities available after training, programmes inevitably create a gap between spending and outcomes. 

The government is now proposing PM-SETU, a ₹60,000 crore programme to upgrade 1,000 government ITIs, with industry-led governance, work exposure, career guidance, and placement support.

PMKVY 4.0 has also introduced future-skill roles, and the wider Skill India ecosystem now covers AI, robotics, cybersecurity, cloud computing, IoT, and data analytics.

The direction is clearly changing. The government is moving from conventional vocational training towards a workforce shaped around emerging technologies. But the central weakness identified in the earlier programme remains relevant: training numbers are not employment numbers.

As Nair puts it, funding and evaluation should follow sustained employment, productivity, and real-world skill application, with live projects and post-certification tracking acting as early warning systems when a skill begins losing market value.