I have read the recent reports on skill development, and like everyone working in this field I have found a great deal in them worth thinking about. This report is my attempt to take the conversation one step further, from describing how work is changing to deciding what a skilling institution should actually build in response.
As machines take on more of the routine cognitive work that can be done at a desk, the contribution that stays distinctly human moves toward the embodied, situated, judgement-laden and accountable work that is the home ground of the polytechnic, the ITI and the skilling institution. India's skilling sector may be unusually well placed for the years ahead, if it builds for the human-AI frontier rather than for the tasks machines are already learning to do.
The pattern underneath the change is, I think, becoming clear. The work that sits closest to the human body, the human hand and human judgement is the work that is hardest to hand to a machine. Robots remain a long way from the everyday physical dexterity a good tradesperson takes for granted, and the practical knowledge they would need to close that distance is on a scale measured not in years but in effective millennia. The opportunity is not to defend human work against the machine but to design for the partnership in which each does what it does best. I call this design principle skilled trade plus AI orchestration, and the longer trajectory it points toward, the steadily leaner enterprise, the Shunya Nigam or Zero-Person Company; both are my own terms, flagged as such wherever they appear.
Three commitments keep this report honest. It takes the optimistic reading seriously, that these tools may lift the judgement of ordinary workers rather than narrow their prospects, and shows the choices it recommends hold up under that reading too. It does not treat work that is hard to automate as automatically work that pays a decent living. And it offers not a fixed prediction but a method you can re-run as the picture changes.
The reallocation: where human value is moving
The argument: what is changing, and what stays human
1. How to Read This Report
Read this as a foresight instrument you can interrogate and re-run, not a forecast to take on faith, and never as a pitch.
Who this is for. This report is written for the people who must commit money and years before the future is fully legible: founders and promoters weighing a 10 to 20 year build, state skill-mission planners working mostly in smaller towns and rural blocks, polytechnic and ITI principals, and skills-university leaders.
How I have handled evidence. Where I state something as fact, it rests on real and publicly available evidence, and a short sources note at the back lets anyone who wishes trace it. Where I am offering my own reading rather than an established finding, I say so plainly. References to the recent body of work on skills are kept general, because the point of this report is to add to that conversation, not to weigh in on any single contribution to it.
A note on terms. Two phrases are my own coinages, flagged at first use. Skilled trade plus AI orchestration is my name for the design principle in Section 8. The Shunya Nigam, or Zero-Person Company, is my term for the steadily leaner enterprise at the far end of this trajectory; I have written on it separately, and Section 8 is explicit about how far I extend that earlier idea here.
One note up front. The central view, that the skilling sector is well placed if it builds for the human-AI frontier, is my own reading, offered for others to test and improve. Section 5 sets the most considered alternative view alongside it before the build chapters proceed.
2. The Foresight Method
Hand a builder a method to read demand, not a list of jobs that dates. The method is the lasting contribution.
A skilling institution is a long promise. The workshop you equip this year will still be teaching a decade from now. The most useful thing a report like this can offer is not a list of the jobs of 2035, which will be out of date before the ink dries, but a method for reading demand that you can run yourself, again and again, as the picture changes.
Read three signals together, not one. The capability signal: what machines can and cannot yet do, read from the gap between controlled demonstration and what holds up in the mess of real work. The demand signal: where the volume of human work is growing, often driven less by technology than by demography. The local signal: what employers within reach of your campus actually hire and pay for, which no global study can tell you. Where the three agree, you can commit; where they disagree, you have found the question worth investigating before you spend.
Grade your sources, and set a cadence. Treat a claim as safe to build on only when it is measured, public and consistent across more than one careful source; treat a striking single number as a lead to verify. Decide in advance how often you will re-read the leading signals, before each tranche of capital. A good institution is not one that guessed the future correctly once. It is one that built itself to keep reading the future as it arrives.
The foresight method: read three signals together
3. The Changing Division of Labour
Machines are taking the codifiable cognitive middle; what remains, and grows, is work that is physical, contextual, relational and accountable.
The useful way to think about what is happening is not as a contest that humans are winning or losing, but as a reallocation. For most of the industrial age, machines took over physical labour and left thinking to people. The current wave runs the other way. The tasks now most readily handed to a machine are cognitive and codifiable: drafting, summarising, classifying, searching, reconciling, the routine production and processing of information that fills so much of a working day at a desk.
It is worth being precise, because the headline version, that knowledge work is finished, is both alarming and wrong. Exposure is uneven: the judgement, the relationship, the accountability and the messy non-standard cases within the same job are far less open to automation than its routine core. It is concentrated in the wealthy, white-collar, desk-heavy economy; work rooted in the physical and the interpersonal is least exposed, and economies weighted toward it, India's among them, sit at the lower-exposure end of every careful estimate. And none of it speaks to the physical world: the evidence that machines are closing on desk work tells us nothing about whether they can wire a building, nurse a patient or repair a pump in a basement that floods.
The shift, then, is real but bounded. For most of the world that is a difficult adjustment. For the skilling sector it is something closer to an opening, because the work that remains is the work it already knows how to teach.
4. The Durable Human Contribution
Four areas stay durably human, the hand, the unstructured situation, the novel case and accountable oversight, with care beside them on demography.
The hand and the body. A skilled tradesperson moving through an unfamiliar site, feeling for a fault, adjusting grip and force to a material that is never quite standard, is performing a feat machines remain a long way from matching. A machine learns physical skill from data about the physical world, and that data exists at nothing like the scale of the text that taught the current systems to write. By credible estimate the shortfall is measured in the equivalent of many thousands of years of practice. This is the firmest ground in the report.
The unstructured situation. The installation that does not match the drawing, the repair under time pressure in conditions no two of which are alike. Demonstrations under controlled conditions are common; sustained, unsupervised performance in the real-world mess is still rare, and the move from one to the other has been consistently slower than its champions expect.
The novel case with no template. The unfamiliar problem, the one-off decision, the situation for which no clean measure of success exists to learn from. Machines are strongest where the past is a good guide to the present and weakest where it is not. That judgement grows more valuable, not less, as the routine cases are handled by machine.
Trust, responsibility and oversight. As more decisions are taken with machine assistance, someone has to remain answerable, and that someone has to be a person. In some jurisdictions regulation now requires a competent human be able to understand, oversee and override an automated system in high-stakes settings; India is developing its own framework for accountability and data responsibility. This turns oversight into a defined, teachable, certifiable role.
Beside these four stands a fifth, working differently. The largest body of durable human work is care: nursing, allied health, support for the elderly and the very young. Its durability rests less on a capability the machine lacks than on a demand rising for reasons of demography, and I treat it in its own right in Section 6.
The durable human frontier: four gaps, plus care
5. The More Optimistic Reading
The build decisions hold whether the cautious or the optimistic economist is right. That is what makes them safe to commit now.
Intellectual honesty requires setting the strongest version of the opposing case beside the argument. The value of these tools may lie less in replacing people than in lifting them: if a machine can put a usable portion of expertise in the hands of someone who has not spent the years to earn it, the effect is to widen entry to well-paid work rather than hollow it. Where these tools have been studied in workplaces, the largest gains have tended to go to the less experienced, narrowing the gap with the best. The economics here are genuinely unsettled, and anyone who tells you the question is settled is selling something.
A builder can still act while the economists argue. If the cautious view holds and routine cognitive work is automated, then the hand, judgement, care and accountability are where durable human work concentrates. If the optimistic view holds and the machine mainly augments, then those who thrive are precisely the people who can pair sound judgement with the tool, the very pairing this report puts at its centre. The two readings disagree about how much routine cognitive work survives; they agree on which human capabilities appreciate in value. That agreement is enough to build on.
The no-regrets test: the same build wins under both futures
6. From Capabilities to Livelihoods
Separate work that survives automation from work that grows on demography, and durable work from work that actually pays.
A frontier of durable capability is not yet a plan. Three clusters carry most of the durable demand. The skilled trades: electrical, plumbing, fitting, installation, maintenance, repair, field and site work, physical, situated and non-standard. Care and the human services: the largest growing pool, rising chiefly because populations are ageing; because it is so heavily done by women, designing for the care cluster and designing for women's access are the same task. Oversight, safety and verification: smaller today, but the cluster most clearly created rather than merely preserved by automation.
Now the hard part. Work that is hard to automate is not the same as work that pays well, and an institution that conflates the two will send its graduates into durability and out of prosperity. Much of the most defensible work, particularly in the trades and care, sits in the informal economy, where pay is low and a qualification is often not recognised. A programme should be judged not only by whether its skills outlast automation but by what its graduates actually earn. What an electrician or a carer earns in a given district is a local and a moving figure that the institution must gather itself; I would rather hand a builder an honest blank than a confident guess.
From durable gaps to clusters to a livelihood
The arrival of capable machines is not a threat to be survived but a partnership to be built well, so that it serves the people it touches.
The build: turning the frontier into an institution
7. The Indian Picture
India already has the scaffolding; the task is to point it at the work that lasts, in a largely informal market.
India has more scaffolding than it is usually given credit for: a national skills framework, a newer credit framework that puts vocational learning on the same ladder as academic learning, a large network of polytechnics, a much larger network of industrial training institutes, sector bodies that define what each job requires, and an apprenticeship system. The question is not whether to invent a system but how to point the existing one at the work that will last.
For any trade you are considering, ask the four questions of Section 4: does it need the hand and the body in non-standard places; does it demand judgement with no template; does it carry responsibility someone must answer for; is its demand driven by a rising human need such as care. The more a trade meets, the more durable it is. Two cautions keep this honest: the precise counts belong in the sources note, confirmed against the current official version; and the defensible trades, the electrician, the plumber, the carer, are precisely the ones most likely in India to work informally, without a formal employer to hire the graduate or honour the credential. In the non-agriculture sector, 73.2 per cent of workers are in informal-sector enterprises (Periodic Labour Force Survey 2023-24). That shapes everything downstream.
8. The Build Principle: Skilled Trade Plus AI Orchestration
Build every programme so the human owns judgement and the hand and the machine owns retrieval and the record, then test that the pairing is real.
This is the hinge on which the report turns from diagnosis to building, and it rests on a single principle, my own term, skilled trade plus AI orchestration: every programme is designed so the human owns the durable parts of the work, the hand, the judgement, the presence and the accountability, while the machine carries the parts it does best, retrieval, planning, documentation, and the two are taught to work as one. Consider an electrician at a fault in an unfamiliar building. The durable work is entirely theirs; beside them a machine holds the manual, surfaces the regulation, suggests a sequence and writes up the job. They are not replaced; they are amplified.
It also tells you what to stop doing: building programmes around the automatable middle, and pouring scarce capital into machines that mimic the human hand. The institution's edge is the human paired with the machine. I connect this to a larger idea I have written about separately, the Shunya Nigam or Zero-Person Company, the steadily leaner firm; I have argued it chiefly in terms of ownership, taxation and governance, and here I extend it to labour, which is my own extension, not a claim of that earlier work: the leaner the enterprise, the more concentrated and valuable the remaining human roles.
The build principle: who owns what, and the test that keeps it honest
9. Curriculum and Pedagogy
Teach dexterity on the awkward, judgement through cases, the machine as a partner, and have the courage to stop teaching the obsolete.
Every major thing a student learns should trace back to a durable area of Section 4, and every trade skill should be taught alongside the machine that now accompanies it. Teach dexterity on the awkward, the worn part, the non-standard fitting, not clean bench tasks a machine could do anyway. Teach judgement through cases, apprenticeship-led, not lecture-and-recall. Pair every module with its machine, and train the student to know when it is wrong, to explain why, and to override it. Assess in the mess, a real or simulated job defended aloud, not a paper test. Credential the human in care, the relational skill, not only the clinical task. And have the courage to stop teaching the obsolete: the clearest sign the principle has been adopted is not what an institution adds, but what it has the discipline to drop.
10. Infrastructure, Faculty and Delivery
Faculty, not buildings, is the binding constraint, and the design must work in the small town, not only the metro.
Build workshops as teaming spaces, not robot showrooms: trade equipment and a modest, reliable digital layer, not humanoid hardware bought to look futuristic. Treat faculty as the binding constraint, because it is, the teaching this report demands can only be done by people who have done the work; bring practitioners in from working life and accept that faculty, not buildings, limits how fast and how well the institution can grow. Put learning inside real work wherever possible. And design for the small town and the village, not the metro: patchier connectivity, thinner faculty supply, lower employer density. The largest growing cluster, care, is overwhelmingly female, so women's access, safe transport, hostels, scheduling, safety, is a design input, not a diversity footnote.
11. The Economics of a Durable Build
The heavier balance sheet is the moat. Fund the durable base as capital and the fast-ageing digital layer as a renewable cost.
The institution this report describes costs more than the alternative, and the capital intensity is the point. A skilling offer that is mostly digital content can be copied, undercut and ultimately produced by the machine itself; an offer built on real equipment, real practitioners and real work cannot be. Spend the capital intelligently: good trade equipment is a long-lived asset, while the digital and machine-intelligence layer ages in a few years. Fund the durable base as long-term capital and the digital layer as a recurring, renewable cost, never as if it were a lathe. The cost can be shared, national skilling funds and state missions, corporate social responsibility, apprenticeship reimbursement, employer co-investment, each mapped to the layer it is best placed to bear. The honest test of the economics is not how many were trained, but what their training was worth to them.
Capital with two lives: do not fund them the same way
12. Credentialing Human Capability
A credential is durable income only if the market reads it, including the informal market with no formal employer.
Credentialing is the mechanism by which durable capability becomes durable income. Use India's credit framework to build stackable competences, so a working person climbs in steps without leaving their livelihood. Two credentials deserve to be created deliberately: one certifying the orchestration skill itself, the trained ability to use, judge and override the machine; one certifying the accountable human authorised to oversee and sign off in a machine-assisted process. The hard version is that for much of India's durable trades there is no formal employer to read a credential. The answers aim it at where trust is actually exchanged: a customer-facing trust mark, recognition on the platforms through which informal work is booked, and the self-recognition routes the credit framework increasingly allows. And track what graduates earn, not merely placements, so the credential is validated by the livelihood it produces.
13. The Build Decision
Commit now to what holds under both futures, stage what depends on local specifics, and re-run the method before each tranche of capital.
Uncertainty is not a reason to wait; it is a reason to commit to what is right under more than one version of the future. Commit now to building for the durable frontier; to making every programme skilled trade plus AI orchestration; to solving faculty first; to designing for the small town, the informal market and women's access; and to measuring success by what graduates earn. Stage the things that depend on specifics not yet settled: the precise mix of trades, the digital layer (bought light and renewed often), and the scale (following the faculty you can secure). Keep watching, and let a meaningful shift in the leading signals, not habit and not a sales cycle, trigger the next tranche of capital.
The build decision under deep uncertainty
The machines will keep getting better at what they do. Our task is to keep getting better at what only we can do, and to build the places that teach it.
Authored terms, the India scaffold and the evidence
A. Glossary of Authored Terms
My term for the steadily leaner, eventually near-zero-headcount firm as the long trajectory of automation. My separate essay frames it in ownership, taxation and governance terms; the labour reading in this report is my own extension here, not a claim of the earlier essay.
The central design principle: the human owns the durable parts of the work, the hand, situated judgement, presence, trust and accountability; the machine carries retrieval, planning and documentation; the two are taught to work as one.
The durable frontier is the set of capabilities that stay human through roughly 2030 to 2035: the hand, the unstructured situation, the novel case, oversight, and care. The exposed middle is the routine, codifiable, desk-bound cognitive work machines are now most readily taking. The no-regrets commitment names the choices that hold under both the cautious and the optimistic reading of the future.
B. Provisional Defensible-Trade Shortlist
A starting point for local investigation, not a finding to act on blind. Trade names are confirmed against the Directorate General of Training Craftsmen Training Scheme catalogue; the mapping to a durable gap is the author's adjudication. India-specific demand and wage data has not been gathered here, so the shortlist is provisional.
| Defensible trade (DGT) | Durable gap it meets |
|---|---|
| Electrician | Hand and body in non-standard settings; accountability |
| Plumber | Hand and body; the unstructured site |
| Fitter / Welder | Dexterity; inspection and safety |
| Mechanic (diesel / motor vehicle) | Field diagnosis under non-template conditions |
| Nursing and allied health | Care; demography-driven demand |
C. The India Scaffold and the Evidence
India's public skilling scaffold, NSQF and the National Credit Framework, NEP 2020 vocational integration, AICTE polytechnics, DGT and the ITIs, NCVET, NSDC and the Sector Skill Councils, and NAPS apprenticeship, gives a builder the system to point at durable work. The figures below were confirmed against primary sources; rupee figures and per-trade wages are deliberately not stated, and some scaffold counts await primary-gazette confirmation.
| Confirmed figure | Source |
|---|---|
| NEP 2020: by 2025, 50% of learners to have vocational exposure (para 16.5); milestone year now passed SAFE | NEP 2020, Ministry of Education |
| 3,865 polytechnics; about 14.1 lakh stand-alone diploma enrolment SAFE | AISHE 2021-22 |
| AICTE Act 1987; Approval Process Handbook 2024-25 to 2026-27 SAFE | AICTE |
| 36 Sector Skill Councils approved SAFE | NSDC |
| 73.2% of non-agriculture workers in informal-sector enterprises SAFE | PLFS 2023-24, MoSPI |
| India has no general human-oversight mandate for automated decisions; accountability via the data fiduciary SAFE | DPDPA 2023, MeitY |
D. Global Benchmark Notes
The vocational systems most often held up as models share one transferable principle: learning held close to real work, defined with the people who do that work, and recognised on equal terms with academic learning. Dual vocational education and training rests on deep employer integration, the limit being its assumption of a dense base of formal employers. The continuous-skilling model treats learning as a lifelong, stackable entitlement coupled to the economy's direction, the limit being scale and a high-coordination setting. Universities of applied sciences hold genuine academic standing while staying firmly applied. The adaptation each asks of an Indian builder is the one this report makes throughout: honour the durable, work-integrated, credit-recognised intent while designing for the informal, dispersed, smaller-town reality the institution will actually live in.
Take the report with you
Download the companion report as a PDF, to read offline or share with your board. RAYSolute advises founders, state skill missions, polytechnics, ITIs and skills universities building for the human-AI frontier.