From Fresher to Supervisor: Why India's IT Sector Must Reinvent Apprenticeship Before the Pipeline Runs Dry

The Vanishing Entry Point

Something structural is happening to entry-level hiring in India's IT sector, and it is being misread as a cyclical correction. IT majors like TCS, Infosys, Wipro, and HCL have all signalled reduced fresher intake over the past two years. The general explanation has been cautious client demand, global headwinds, and post-pandemic normalisation. That explanation is partially true and largely misleading.

The deeper driver is AI absorbing the cognitive work that entry-level IT roles were built around — basic coding, data processing, initial testing, routine documentation, first-pass debugging. These tasks have not disappeared; they have been reassigned. The machine now does them faster, cheaper, and without the onboarding costs. From a company's quarterly perspective, this looks like efficiency. From the sector's five-year perspective, it is the beginning of a pipeline problem.

Why This Is Not Just a Jobs Problem

The instinct is to frame vanishing entry-level roles as an employment crisis — and it is that, but it is also something more structurally serious. Entry-level positions in Indian IT were never merely cheap labour. They were the apprenticeship layer through which domain judgement, error intuition, and supervisory competence got built. The associate developer who spent two years debugging code developed an instinct for where systems fail. The junior analyst processing data pipelines developed a feel for anomalies that no classroom produces. That tacit, embodied knowledge is what eventually made them credible senior professionals.

A recent research report by Anthropic, titled "Agentic Coding and Persistent Returns to Expertise", published on 16 June, confirms this with empirical precision. The study, which analysed roughly 400,000 AI-assisted coding and work sessions, finds that domain expertise is the decisive variable in human-AI collaboration — the more expertise a person brings, the more productive the AI becomes in their hands. Critically, the largest expertise gap is not between intermediate and expert users, but between novice and intermediate. The highest-leverage investment, the report says, is lifting people from knowing nothing to knowing enough — which is exactly what entry-level operational experience used to accomplish. Remove that experience layer, and you do not just lose those jobs, you interrupt the formation of the judgement that the AI-era IT sector will depend on.

The Supervisory Economy Is Coming — But Who Will Staff It?

The demand side of the future IT labour market is already visible. As AI handles routine and mid-level coding, testing, and analytics tasks, it generates a new category of roles: AI workflow monitors, code review supervisors, agent orchestrators, and human-in-the-loop validators who catch what machines miss, steer outputs toward contextual accuracy, and bear accountability for consequential technical decisions. These are not marginal roles. In enterprise IT — where a hallucinated compliance output, a flawed architecture decision, or a biased algorithm carries real operational and reputational consequences — credible human oversight is not optional. It is the load-bearing layer of the entire AI-augmented operation.

The human-in-the-loop (HITL) economy assumes a supply of humans capable of that oversight. But that supply does not self-generate. It is produced by a pipeline — and the pipeline runs through the entry-level operational experience that AI is now absorbing. India's IT sector is simultaneously creating the demand for supervisory talent and destroying the mechanism that produces it. That is the structural contradiction at the heart of the current moment.

Why Formal Education Alone Cannot Bridge the Gap

The obvious response is to fix colleges/universities — redesign curriculums, introduce AI literacy, add ethics modules, build interdisciplinary engineering programs. All of this is necessary. None of it is sufficient.

Formal education can transmit conceptual knowledge: what a hallucination is, how bias enters a model, what responsible AI governance frameworks require, how software architecture decisions cascade. What it cannot replicate is the tacit knowledge built through operational engagement — the feel for when an AI-generated output is plausible but wrong, the instinct that flags anomalies before they can be articulated, the judgment about when to intervene and when to trust. The Anthropic report is explicit on this point: participants who fully delegated coding tasks to AI showed some productivity gains but demonstrably failed to learn the underlying domain. Delegation without engagement produces efficient operators, not credible supervisors.

The medical profession solved an analogous problem through residency — a structured post-graduation period where doctors work in real institutional settings, on realistic cases, under supervision, before they are credentialed as independent practitioners. 

The IT profession now faces the same design challenge. The answer is not more classroom instruction. It is a new kind of apprenticeship.

The New Apprenticeship — Design Principles

The proposal is specific. Graduates equipped with T-shaped education — defined by deep technical domain expertise paired with broad interdisciplinary competence — should be placed in IT companies to work alongside deployed AI systems. Not to supervise the AI, which requires senior judgment they have not yet developed. But to work with it: to engage with real or realistically simulated enterprise software workflows, to process AI-generated code and outputs critically, to learn through repetition and consequence where AI performs well and where it drifts.

The T-shaped profile is not incidental to this design — it is foundational. The vertical bar of deep technical knowledge is what makes oversight eventually credible. A graduate placed in an AI-assisted software development workflow without understanding software architecture, testing logic, or systems design cannot distinguish a plausible error from a correct output. The horizontal bar — project management capability, awareness of organisational and societal impact, ethical reasoning about algorithmic consequences — is what makes oversight consequential rather than merely technical. Together, these two bars define a talent profile that neither pure engineering education nor pure management education currently produces. The apprenticeship is where the T-shape becomes operational.

Several design principles are non-negotiable. Simulation fidelity must be high — toy problems produce toy competence. The work must mirror actual enterprise AI deployments, with real ambiguity and real error profiles. The program must be time-bound and structured, with clear milestones and progressive responsibility. And the credential it produces must be portable — recognised across the IT sector, not proprietary to the company that ran the program. A company-specific certificate is a retention tool. A sector-wide credential is a public good.

The Institutional Architecture — NASSCOM, NSDC, NAPS 

India already has the administrative machinery to run this at scale. The National Apprenticeship Promotion Scheme has demonstrated that it can mobilise employer participation and stipend support — the manufacturing sector's recent robust uptake is proof of concept. Engineering CPSUs have run apprenticeship programs for decades. The challenge is extending this architecture to white-collar, knowledge-workflow contexts where the tool is an AI system rather than industrial equipment.

The institutional design should be association-led, not company-led. Three bodies need to play distinct, non-overlapping roles.

NASSCOM is the anchor and convener. As the representative body of India's IT industry, it alone has the sectoral authority to define what HITL and AI supervisory competence actually means in enterprise software contexts. NASSCOM's specific responsibilities would be: co-designing the apprenticeship curriculum with input from large IT employers; defining the competency framework against which apprentices are assessed; accrediting companies as eligible apprenticeship hosts; and ensuring the resulting credential reflects real industry requirements rather than generic AI literacy. Without NASSCOM's ownership of curriculum design, the program risks being government-designed training that the industry neither trusts nor values.

NSDC — the National Skill Development Corporation — provides the qualification architecture. Its role is to formally recognise the HITL apprenticeship credential within India's National Skills Qualifications Framework, making it portable, nationally legible, and stackable with other qualifications. NSDC also brings experience in scaling sector skill councils and can help NASSCOM institutionalise what is initially a pilot into a durable, replicable program structure.

NAPS — the National Apprenticeship Promotion Scheme — is the funding and compliance engine. It reimburses a stipend share to employers who take on registered apprentices, handles employer registration and compliance tracking, and provides the legal and administrative scaffolding within which the program operates. Without NAPS, the financial burden of running apprenticeship cohorts falls entirely on firms whose short-term incentive to do so is weak. With NAPS co-financing, the calculus changes — particularly for mid-tier IT companies that lack the deep pockets of a TCS or an Infosys but have equally urgent long-term need for supervisory talent.

Together, the three bodies cover the full institutional stack: NASSCOM owns domain credibility and curriculum; NSDC owns qualification legitimacy and scalability; NAPS owns financing and administration. No single body can substitute for the others. The partnership is not a convenience — it is a structural necessity.

Large Indian IT companies — TCS, Infosys, Wipro, HCL, Tech Mahindra — should form the first cohort of employer partners. They are large enough to design credible simulation environments, sophisticated enough to understand what HITL competence actually requires, and exposed enough to AI disruption pressure to have genuine long-term interest in building supervisory talent pipelines. Their participation would also signal to mid-tier IT companies that the program is industry-endorsed rather than government-imposed.

The Message to NASSCOM

NASSCOM has invested considerable energy in framing AI as a net opportunity for India's IT workforce — new roles, new capabilities, new global positioning. That framing is not wrong. But it is incomplete without the pipeline architecture that makes it durable.

The supervisory roles NASSCOM is taking for granted requires a steady supply of IT professionals who understand AI systems deeply enough to oversee them credibly. That supply does not appear spontaneously from engineering colleges. It is built through structured engagement with deployed AI over time — exactly what the new apprenticeship model provides. The proposal here is not a challenge to NASSCOM's optimism. It is the infrastructure that makes the optimism sustainable.

NASSCOM needs to see the problem before it can lead the solution. The hiring contraction at the entry level is visible now. The HITL talent shortage it will produce is five to seven years away. Institutions do not typically act on problems that have not yet arrived. This is an argument that the window for proactive design is open — but it will not remain open indefinitely.

Conclusion: Act for Tomorrow, Today 

India's IT sector has the institutional actors and the policy precedents to build a new apprenticeship architecture for the AI era. NASSCOM, NSDC, and NAPS are all already on the ground. What is missing is the diagnosis that connects the entry-level hiring contraction of today to the supervisory talent shortage of tomorrow — and the institutional will to act on that connection before the gap becomes a crisis.

The pipeline takes years to build. The demand for what it produces is already forming. The time to begin is not when India's IT majors are scrambling for HITL talent they cannot find. The time to begin is now.