The Portable Worker, The Intelligent State, AI, Administrative Modernisation, and India’s Quest for Seamless Social Security
For decades, India’s social security architecture was designed for a world that no longer exists. It was a world of permanent employment, single careers, and geographic stability—a world in which a worker joined a factory or an office in their twenties, remained with the same employer for forty years, and retired with a pension from the same provident fund account they had opened on their first day of work. In that world, social security could be organised around the employer, the establishment, the permanent contract. The worker’s identity was stable because the worker’s life was stable.
That world has receded into memory. Nearly 90 per cent of India’s workforce now operates in informal employment arrangements, a category that encompasses not only the traditional informal sector of daily wagers and self-employed artisans but also the rapidly growing workforce of gig workers, platform labourers, and freelance professionals who move fluidly between employers, occupations, and states. For these workers, a single year may include multiple employment spells: three months driving for a ride-hailing platform, two months delivering food for an aggregator, six months on a fixed-term contract with a manufacturing firm, a month of freelancing on a digital marketplace. Each spell generates contributions to some scheme or combination of schemes. Each scheme maintains its own ledger. And when the worker moves on, the ledger is archived and effectively forgotten.
The Social Security Code, 2020, and its accompanying draft rules represent a deliberate and long-overdue attempt to correct this mismatch. By formally recognising gig workers, platform labourers, and the unorganised sector within a unified legal framework, the Code shifts the conceptual anchor of social security from the contract to the worker. It acknowledges that the only constant in contemporary Indian employment is the worker herself; the system must therefore organise itself around her, not around the ephemeral and fragmented jobs she performs.
This is a transformative legislative intent. But intent is not implementation. The task now confronting India’s social security administration is the translation of legal recognition into lived protection for approximately 94 crore beneficiaries—a population larger than that of any country except India itself. This task is not merely administrative; it is conceptual, technological, and federal. It requires reimagining the relationship between citizens and the state, between data and rights, and between the Union government and the States. And it requires, at its core, a capability that India has already demonstrated at population scale but must now deploy with unprecedented sophistication: the capacity to track, connect, and serve individuals as they move across the vast, complex terrain of India’s labour markets.
The Portability Imperative: From Fragments to Continuity
The defining feature of India’s contemporary labour market is mobility. Workers do not merely change jobs; they change occupations, sectors, and states. A construction worker in Maharashtra may return to his village in Uttar Pradesh for the harvest season and then migrate to Delhi for work in the gig economy. A young professional may spend three years in formal employment covered by the Employees’ Provident Fund Organisation (EPFO), then shift to freelance work registered on the e-Shram platform, then accept a fixed-term contract with a startup that contributes to the Employees’ State Insurance Corporation (ESIC). Each transition is rational from the worker’s perspective; each transition is also an administrative rupture.
India’s social security systems evolved in silos. EPFO, ESIC, e-Shram, state welfare boards, and scores of other schemes each maintain their own beneficiary registries, contribution records, and eligibility criteria. When a worker moves from one scheme to another, her prior contribution history is not automatically recognised. She must register anew, establish her identity afresh, and begin accumulating contributions from zero. Her prior contributions are not lost—they remain in the ledger of the previous scheme—but they are functionally inaccessible for the purpose of establishing eligibility for benefits under the new scheme. The worker experiences this as a reset; the system experiences it as the normal operation of fragmented, scheme-specific administration.
This is the portability deficit, and it is the central challenge that the Social Security Code must address. The Code’s recognition of the worker as the unit of social security is the legal foundation for portability; what remains is the operational infrastructure that would enable contributions and benefits to move seamlessly with the worker across schemes, sectors, and states.
That infrastructure has three components. First, a universal, persistent worker identifier that can link employment and contribution records across schemes. India already possesses this in the Universal Account Number (UAN) under EPFO, which has been extended into platforms like e-Shram. The technical foundation exists; what is required is its consistent application across all social security schemes. Second, interoperable data standards that enable different systems to exchange information about worker identities, contribution histories, and eligibility status. This is a governance challenge, not a technical one; the standards exist, but their adoption requires coordination and compulsion. Third, and most critically, an intelligent connective layer that can perform the linking—that can recognise, in real time or near-real time, that the worker who has just registered for e-Shram is the same worker who contributed to EPFO for three years, and that those contributions should be aggregated with future contributions to determine her eligibility for benefits.
This third component is where artificial intelligence becomes relevant as a governance capability.
The AI Opportunity: Connection, Not Prediction
The authors of the accompanying article—senior officials and technologists with direct responsibility for India’s social security systems—offer a crucial reframing of AI’s role in governance. “The value of AI,” they write, “lies not in prediction alone, but in connection.” This is not the AI of science fiction, autonomously deciding who deserves benefits. It is the AI of pattern recognition, entity resolution, and process optimisation—applied not to replace human judgement but to enable systems that currently operate in isolation to function as an integrated network.
Consider the concrete use cases. When a worker moves from construction employment in Maharashtra to platform-based gig work in Karnataka, her EPFO contributions cease and her e-Shram registration begins. In the current fragmented system, this transition is invisible to the administration. The worker’s cumulative contribution history is interrupted; her eligibility for benefits that require continuous contribution periods may be jeopardised; her entitlement to portability—theoretically guaranteed—is practically denied.
An AI system trained on historical patterns of employment transitions can identify this movement with high confidence. It can flag the contribution lapse not as an enforcement failure but as a transition event, triggering automated verification of the worker’s new employment status and seamless linkage of her pre-transition and post-transition contribution records. It can generate a personalised notification to the worker, explaining that her social security coverage has been maintained and advising her of any additional steps required. It can provide the Karnataka gig platform with verified information about the worker’s prior contributions, reducing their compliance burden and incentivising formal registration.
This is not speculative. Every component of this workflow already exists in India’s digital infrastructure. The UAN provides the persistent identifier. Aadhaar provides the authentication. EPFO and e-Shram provide the transactional data. What is missing is the intelligent orchestration layer that connects these components dynamically, in response to worker mobility rather than static administrative schedules.
The same connective intelligence can be applied to other dimensions of social security administration. Compliance monitoring can shift from random inspections to risk-based targeting, focusing enforcement resources on establishments where pattern analysis suggests systematic underpayment or evasion. Eligibility determination can become proactive rather than reactive, identifying workers who are approaching qualification thresholds and notifying them of their impending entitlements. Grievance resolution can be accelerated through automated triage that routes cases to the appropriate authorities based on the nature of the complaint and the worker’s employment history.
These applications share a common characteristic: they enhance rather than replace human judgement. They do not automate decisions; they inform them. They do not eliminate discretion; they discipline it with data. They do not centralise control; they enable coordination across federal partners. This is AI as institutional infrastructure, not AI as autonomous agent.
The Federal Dimension: Strengthening States Without Centralising Control
India’s federal structure, with labour on the Concurrent List, adds complexity to any unified social security architecture. States are not merely implementing agencies; they are constitutional partners in the delivery of welfare, with their own priorities, capacities, and political accountabilities. Any system that appears to centralise control or displace State autonomy will meet resistance, regardless of its technical sophistication.
The authors’ framing of AI’s federal role is therefore strategically significant. They emphasise that national digital systems should provide “common foundations through shared identifiers and interoperable standards,” while States remain “closest to workers and employers, shaping inspections, enforcement priorities, and welfare delivery.” AI, in this conception, is not an instrument of centralisation but an enabler of cooperative federalism. It allows States to retain their distinctive approaches to welfare delivery while participating in a national ecosystem of portable entitlements.
A worker registered with the Kerala Welfare Fund for construction workers who moves to Tamil Nadu does not need to surrender her Kerala entitlements and begin anew. The shared identifiers and interoperable standards enable Tamil Nadu’s systems to recognise her prior contributions, apply State-specific eligibility rules, and ensure continuity of protection. The worker experiences seamless portability; the States retain policy autonomy; the national system provides connective infrastructure without imposing substantive uniformity.
This is the institutional role for AI that the authors advocate: not a replacement for federal negotiation but a tool that makes federalism work better by reducing the transaction costs of inter-State cooperation. It is a vision of technology as constitutional lubricant rather than constitutional disruptor.
The Cautionary Note: Digitisation Is Not Reform
The authors are emphatic, however, that AI is “not a shortcut to universal social security.” They draw on hard-won lessons from earlier phases of digitisation, which too often consisted of layering new technology over old pathologies. Digitising a poorly designed procedure does not fix it; it amplifies its inefficiencies at scale. AI layered over unclear roles, inconsistent data, or weak accountability will not produce intelligent governance; it will produce intelligently automated dysfunction.
The sequencing imperative is therefore critical. Initial AI deployments should focus on low-risk, high-impact use cases that build confidence and demonstrate value without exposing vulnerable populations to algorithmic error. Cleaning registries of duplicate or dormant entries. Clearing backlogs of unprocessed claims. Detecting compliance gaps through pattern analysis rather than random inspection. Reducing the repetitive administrative work that consumes frontline officials’ time and energy.
Only once these foundational capabilities are in place should AI be extended to more consequential domains—and even then, only within sandboxed environments where errors can be contained and corrected without systemic harm. The authors’ caution is not Luddism; it is the prudence of practitioners who have seen technology fail when divorced from institutional context.
This caution extends to the design of AI systems themselves. Algorithms trained on historical data will reproduce historical biases; if past compliance enforcement has been skewed against certain sectors or regions, AI risk models will perpetuate that skew. If eligibility determinations have historically favoured male workers in formal employment, AI tools trained on those determinations will continue to disadvantage women, gig workers, and informal labour. Addressing these biases requires not merely technical fixes but deliberate attention to equity in the design and deployment of AI systems—attention that is currently more aspiration than practice.
Trust as Infrastructure
The authors’ concluding emphasis on trust is not rhetorical flourish; it is operational necessity. Employers need predictability in their compliance obligations. Workers and unions expect transparency in how their data is used and their entitlements determined. State governments must see national digital systems as enabling rather than intrusive. Each of these constituencies has legitimate concerns that must be addressed through visible, verifiable system design.
AI systems built with shared dashboards, common standards, and clear audit trails can reduce discretion and its attendant risks of arbitrariness and corruption. When eligibility determinations are based on transparent rules applied consistently across cases, when contribution histories are verifiable by workers themselves, when enforcement priorities are derived from objective data rather than subjective suspicion—in each of these dimensions, AI can strengthen rather than weaken the trust that sustains large-scale welfare systems.
But this outcome is not automatic. It requires deliberate choices: to prioritise transparency over proprietary secrecy, to design for auditability rather than black-box optimisation, to subject algorithmic systems to the same constitutional scrutiny applied to human decision-makers. The technology does not determine its own governance; governance determines the technology’s impact.
Conclusion: From Legal Design to Lived Reality
The Social Security Code embodies a clear and welcome statement of intent: that India’s social protection system will reflect how Indians actually work, not how they worked in the mid-20th century. The expansion of coverage from 19 per cent to 64 per cent of the population over the past decade demonstrates that India can scale social protection when it commits to doing so. The digital infrastructure—UAN, e-Shram, Aadhaar, India Stack—provides the foundational layers on which portable, worker-centric social security can be built.
What remains is the execution challenge: converting this intent, this demonstrated capacity, and this infrastructure into seamless, reliable protection for the 94 crore workers who currently hold entitlements in fragmented, disconnected silos. This is not primarily a technological challenge; the technology exists, and India has already demonstrated mastery of population-scale digital systems. It is an institutional challenge: reconfiguring administrative routines that have hardened over decades, aligning incentives across federal partners, and building the trust that enables workers to claim their rights and employers to fulfil their obligations.
AI, deployed judiciously and anchored in institutional strength, can help meet this challenge. It can connect systems that currently operate in isolation, identify transitions that currently go unremarked, and reduce the transaction costs that currently exclude mobile workers from the protection they have earned. It can make government not merely efficient but deliberately human—responsive to the actual rhythms of working life rather than the administrative categories of a vanished era.
This is how India can show the way. Not through technological triumphalism, treating AI as a magical solution to deep-seated structural problems. But through institutional pragmatism: identifying where technology can enable human judgement, where data can illuminate policy choices, and where digital infrastructure can strengthen rather than supplant the federal compact. The Social Security Code offers the legal framework. India’s administrative systems offer the operational scale. The task now is to bring them together in a system that travels with the worker, from job to job, from State to State, from vulnerability to security.
That is the promise of portable social security. And that is the work ahead.
Q&A Section
Q1: What is the “portability deficit” in India’s current social security architecture, and why does it disproportionately affect informal and gig workers?
A1: The portability deficit refers to the fragmentation of a worker’s social security history when they move between schemes, employers, or jurisdictions. Because India’s social security systems evolved in silos—EPFO, ESIC, e-Shram, state welfare boards, each with its own registry, contribution records, and eligibility criteria—a worker’s prior contributions are not automatically recognised when they register with a new scheme. The worker must start from zero each time, even though they have contributed continuously across their working life. This disproportionately affects informal and gig workers because mobility is the defining feature of their employment. A gig worker may work for multiple platforms in a single year; a construction worker may migrate seasonally between states; a freelance professional may shift between formal employment and independent contracting. Each transition triggers an administrative reset. For a worker in traditional formal employment (one employer, forty years, single EPFO account), the portability deficit is irrelevant. For the 90 per cent of India’s workforce in informal arrangements, it is a systematic mechanism of exclusion. Their contributions are recorded but not aggregated; their eligibility for benefits is perpetually reset; their entitlement to social security is theoretical, not operational.
Q2: How does the article reframe the value of AI in governance, and why is this reframing significant for the social security context?
A2: The article reframes AI’s value from prediction to connection. In popular discourse and much policy discussion, AI is associated with forecasting—predicting which workers will quit, which firms will default, which regions will experience unemployment. The authors, all practitioners, argue that AI’s more immediate and consequential value lies in connecting existing systems that currently operate in isolation. This reframing is significant for several reasons. First, it democratises AI: it does not require exotic capabilities, massive computational infrastructure, or the replacement of human decision-makers. It requires pattern recognition, entity resolution, and process optimisation—applied not to speculate about the future but to make the present system function coherently. Second, it positions AI as a tool for institutional integration rather than institutional disruption. The goal is not to replace EPFO, ESIC, and e-Shram with a single unified system—a politically and administratively impossible task—but to enable them to function as an integrated network while remaining organisationally distinct. Third, it reframes the AI challenge from technical to institutional. The hardest problems are not algorithmic; they are problems of data standards, inter-agency coordination, federal cooperation, and trust. This reframing is a prerequisite for realistic, effective deployment of AI in social security administration.
Q3: What are the “clear lessons from earlier phases of digitisation” that the authors caution must be heeded, and how do these lessons shape their proposed sequencing?
A3: The authors identify two clear lessons. First, avoid digitising poorly designed procedures. Layering technology over unclear roles, inconsistent data, or weak accountability does not fix these underlying problems; it amplifies inefficiencies at scale. Business process redesign, rule clarification, and data cleaning must precede technological intervention. Second, move beyond administrative silos. AI draws its value from identifying patterns across systems. If systems remain deliberately disconnected—if EPFO, ESIC, e-Shram, and state welfare boards continue to operate as independent fiefdoms—AI cannot perform its connective function.
These lessons shape the authors’ proposed sequencing. Initial efforts should focus on low-risk, high-impact use cases that build confidence and demonstrate value without exposing vulnerable populations to algorithmic error: cleaning registries of duplicate or dormant entries, clearing backlogs of unprocessed claims, detecting compliance gaps through pattern analysis, reducing repetitive administrative work. Only once these foundational capabilities are in place should AI be extended to more consequential domains—and even then, only within sandboxed environments (such as the AIKosh platform) where errors can be contained and corrected without systemic harm. This sequencing reflects the prudence of practitioners who have seen technology fail when divorced from institutional context.
Q4: How does the article conceptualise AI’s role within India’s federal structure, and why is this conceptualisation strategically important?
A4: The article conceptualises AI as an enabler of cooperative federalism, not an instrument of centralisation. National digital systems provide “common foundations”—shared identifiers (UAN, Aadhaar), interoperable standards, and connective infrastructure—while States retain autonomy over inspections, enforcement priorities, and welfare delivery. AI enables this division of labour by reducing the transaction costs of inter-State coordination. A worker moving from Kerala to Tamil Nadu can have her prior contributions recognised by Tamil Nadu’s systems without requiring either State to surrender its policy autonomy.
This conceptualisation is strategically important because it addresses the legitimate concerns of State governments that national digital systems might be mechanisms of encroachment on their constitutional domain. Labour is on the Concurrent List; States are not merely implementing agencies but constitutional partners with their own priorities, capacities, and political accountabilities. Any system that appears to centralise control or displace State autonomy will meet resistance, regardless of its technical sophistication. By framing AI as a tool that strengthens State capacity without centralising control, the authors position technological reform as compatible with, indeed supportive of, India’s federal compact. The worker experiences seamless portability; the States retain policy autonomy; the national system provides connective infrastructure without imposing substantive uniformity. This is AI as constitutional lubricant, not constitutional disruptor.
Q5: What does the article mean by describing AI as making government “deliberately human,” and how does this phrase capture the authors’ broader philosophy of technological reform?
A5: The phrase “deliberately human” captures three interconnected propositions about the relationship between technology and governance. First, that the purpose of AI is to enhance, not replace, human judgement. AI systems should inform decisions, not automate them; they should provide officials with better information, analysis, and recommendations, but the ultimate responsibility for decisions affecting workers’ lives must remain with accountable human beings. Second, that the design of AI systems should be responsive to the actual rhythms and realities of human work. A system that tracks employment transitions, flags contribution lapses, and identifies eligibility gaps is “deliberately human” because it is designed around how workers actually move through the labour market, not around the administrative categories of a vanished era of permanent, single-employer employment. Third, that the deployment of AI should be guided by deliberate, transparent choices about values and trade-offs. Which biases in historical data need to be corrected? Which forms of discretion should be preserved and which constrained? How should conflicts between efficiency and equity be resolved? These are not technical questions to be answered by algorithms; they are human questions to be answered through democratic deliberation and accountable decision-making. The phrase thus captures the authors’ broader philosophy: technology is not an autonomous force that will inevitably transform governance in predetermined directions. It is a tool whose impacts are determined by the choices we make in designing, deploying, and governing it. Making government “deliberately human” means making those choices deliberately, transparently, and accountably.
