Arya Roy Bardhan
NEW DELHI: Artificial intelligence (AI) can no longer be treated as just another line item in the Union Budget. AI is better understood as a general-purpose technology: it can reshape many sectors, but only when complementary investments in skills, data, organisational change, and innovation arrive alongside it.
That is a core lesson from the general-purpose technology literature, namely that the big gains come from “innovational complementarities”, not from the tool alone.
Complementarity I: Human capital
The first implication of AI concerns human capital. AI will not only reward people who can build models, but also raise returns for workers who can use AI tools in ordinary jobs.
For this reason, AI is likely to affect a very large share of jobs, with both productivity gains and displacement risks, depending on how economies adapt. The task-based view of technology is crucial here: automation can displace tasks, but new tasks can also emerge.
Policy influences this balance by shaping skills, mobility, and adoption incentives. In practical Budget terms, the question is whether India is financing “AI-capable work” at scale through digital literacy, foundational numeracy, domain training (such as in health, manufacturing, and logistics), and short-cycle credentials that help workers integrate AI into workflows.
AI is likely to affect a very large share of jobs, with both productivity gains and displacement risks, depending on how economies adapt.
This is where current allocations reflect both intent and gaps. The Budget’s AI-facing provisioning is evident, but still modest relative to the breadth of the transition. For 2025-26, the IndiaAI Mission was budgeted at Rs 2,000 crore, alongside Rs 1,250 crore for research and development (R&D) in IT, electronics, and cyber security, communications, and big data technologies (CCBT), Rs 7,000 crore for the modified semiconductors and display ecosystem scheme, and Rs 9,000 crore for the Ministry of Electronics and Information Technology’s (MeitY) production-linked incentive (PLI).
Finance Minister must move beyond AI outlays to fund upskilling
The National Apprenticeship Training Scheme (NATS) was budgeted at Rs 1,178 crore. This bundle maps to a sensible pipeline: from capability building (IndiaAI / R&D) and hardware and supply capacity (semiconductors / PLI), to limited labour-market linkage (apprenticeships).
However, it still leaves the hardest human-capital problem under-addressed, namely, mass adaptation in non-AI occupations.
International evidence suggests why. Even when AI exposure is high, most workers do not need specialised AI engineering skills – instead, they need a reshaped mix of cognitive, managerial, and digital skills because tasks and work organisation change.
This points to a clear next step for the upcoming Budget. It must treat apprenticeships and skilling as the main delivery channel for AI diffusion in the labour market, not as a side programme.
A practical anchor already exists in public policy design. IndiaAI’s Cabinet-approved mission architecture explicitly includes a “FutureSkills” pillar, with plans to expand AI courses and set up data and AI labs beyond tier-1 cities.
The Budget can now translate this architecture into scaled targets. The outcomes should be measured by the number of apprentices trained on AI-enabled shop floor tools, the number of teachers trained in AI-assisted pedagogy, and the number of micro, small and medium enterprise (MSME) workers credentialed in AI-enabled quality control or logistics. This will ensure that human capital spending is measurable, rather than merely announced.
Complementarity II: Productivity
The second implication of AI concerns productivity. AI’s productivity effects often arrive with a lag because firms must first invest in implementation, including data pipelines, workflow redesign, governance, and complementary intangible capital.
This is the modern productivity paradox argument — capabilities can look impressive, while measured productivity stays muted until diffusion and reorganisation catch up. That framing should shape how we read AI allocations.
Hardware incentives and R&D are necessary, but the Budget also needs an adoption strategy that reaches ordinary firms. Otherwise, AI deployment becomes concentrated in a few large companies and a few use-cases, while aggregate productivity gains remain limited.
AI’s productivity effects often arrive with a lag because firms must first invest in implementation, including data pipelines, workflow redesign, governance, and complementary intangible capital.
The missing bridge is typically finance and standards for adoption. IndiaAI’s Cabinet note is explicit about “democratising computing access,” building a public AI compute capacity of 10,000+ GPUs through public-private partnerships (PPP), and creating a marketplace for AI-as-a-service and pre-trained models.
This is a productivity lever because it lowers fixed costs for startups and smaller firms. However, the Budget still needs to tie that supply-side effort to demand-side uptake.
Credible procurement pathways, sandboxes for regulated sectors, and public-sector demonstration projects that create repeatable templates are necessary. For example, AI-supported diagnostics in district hospitals or AI-enabled inspections in customs and logistics could be introduced.
The aim should be to convert compute availability into adoption at the frontier and, crucially, adoption in the middle of the distribution, where most employment sits.
Complementarity III: Growth
The third implication concerns growth. Over the medium term, the growth payoff depends on whether AI raises total factor productivity broadly while keeping macro risks contained, including labour-market disruption, market concentration, and uneven readiness across states and sectors.
One useful way to organise this is the IMF’s AI Preparedness Index framework, which treats preparedness as a bundle of digital infrastructure, human capital and labour policies, innovation capacity, and legal frameworks.
This matters for budgeting because it suggests the need for balanced inputs, as over-funding one pillar does not compensate for a missing pillar. A compute stack without skills slows diffusion; skills without access to tools limit productivity; innovation without trustworthy governance can raise adoption frictions.
A compute stack without skills slows diffusion; skills without access to tools limit productivity; innovation without trustworthy governance can raise adoption frictions.
In this context, India’s Budget choices show a recognisable direction, but the next Budget needs sharper prioritisation. India has created fiscal space for innovation through a large new Research, Development, and Innovation (RDI) scheme (Rs 20,000 crore in 2025-26) and continues funding for Skill India (Rs 2,700 crore).
These are large levers, but they are also broad. The opportunity is to ring-fence an “AI translation” component within them: finance that moves AI from the lab to the shop floor and from pilots to adoption, with clear outcome metrics. Without that, AI spending can remain fragmented, failing to cohere into a single growth strategy.
________________________________________
Arya Roy Bardhan is a Junior Fellow with the Centre for New Economic Diplomacy at the Observer Research Foundation