Enterprise DPI can help India reclaim its advantage in the AI game

How enterprise digital public infrastructure can become the new rails on which India re-enters the AI game

Strategy is the reconciliation of unlimited aspirations with limited resources. As artificial intelligence (AI) hype collides with a revised assessment of large language model defensibility and the equity returns of data centre investments, India has a chance to re-enter the AI game with a strategy that leverages its unique digital public infrastructure (DPI). This strategy must include building a DPI suite for enterprises: Unique and verifiable identity, entity DigiLocker, moving government interfaces from PDFs to application programming interfaces (APIs), and a single digital source of truth for laws (India Code). This will expand our formal enterprise population and enable evidence-based policymaking. 

India’s strategy for mass prosperity needs a renegotiation of our government’s uneasy relationship with enterprises: 45 per cent of our labour force still works in agriculture because our 70 million-plus employers translate into only 30,000 companies that have paid-up capital exceeding ₹10 crore. Raising manufacturing employment (11 per cent of workers), corporate research spending (1 per cent of gross domestic product), and formal employment (20 per cent of workers) needs more high-productivity employers. Reducing regulatory cholesterol is key, but this also needs digitising the interface, workflows, and relationships between government and enterprises. 

Enterprise DPI will provide iterative policy evidence. In 2020, an American Economic Review research paper, “Rural Roads and Local Economic Development,” used night-time satellite light intensity as a proxy for local economic output and suggested that India’s $40 billion (about ₹3.2 trillion) rural roads programme produced no detectable gains in village incomes or consumption. But a new working paper tentatively titled “Public Infrastructure for Measuring Welfare”, presented by co-author Ravi Bapna of the University of Minnesota at a recent National Council of Applied Economic Research conference on AI and DPI, suggested that payment data was better at predicting the consumption effects of infrastructure than night-time satellite data. The paper argues that data from population-scale fintech PhonePe (700+ million users and 50+ million merchants) shows that 52 per cent of the information signal from payments comes from changes in economic activity within a district over time, compared with just 3 per cent for night-time lights. 

In other words, flow, not stock. Imagine the policy impact of a research annuity on the digital exhaust of 1 billion active internet users (up from 622 million six years ago), all using enterprise DPI to transact at data costs of ₹15 per gigabyte (a tenth of the global average). 

Enterprise DPI is less a digital moonshot and more common sense catching up with technology. It will replace dozens of disconnected regulatory systems with a federated architecture in which identities, licences, certificates, payments, and compliance move seamlessly across platforms. Instead of spending up to a quarter of a sole proprietor’s time navigating portals and paperwork, compliance should become automated, agent-enabled, interoperable and embedded into business software. The same enterprise DPI building blocks — identity, trusted data sharing and payments — will simultaneously improve access to credit, make formalisation attractive, and improve trust with customers and employees. Let’s take a closer look at each pillar. 

Unique & verifiable enterprise identity: Economist Ronald Coase’s argument that firms exist to reduce transaction costs has been lost in government interfaces. Today, a typical enterprise juggles at least 25 identification numbers — roughly 16 from the Union government and nine from states. Each serves a purpose, but together they resemble a bureaucratic Sudoku puzzle. A unique enterprise identifier, anchored in PAN 2.0, can replace this alphabet soup with a pervasive machine-readable identity for all types of enterprises, with an eKYC applicable to myriad uses. It gives AI agents trusted rails to transact securely, enabling entity payments with the ease of UPI. 

Entity DigiLocker: Enterprise documents today are scattered across portals, inboxes, and departments, with no unified way to access or share them. The scale explains the chaos: Union departments issue about 489 types of enterprise documents, while states add over 3,000. EntityLocker can become the default interface for enterprise documents by enabling secure, real-time sharing of licences, registrations, and approvals directly from source departments. Instead of repeatedly uploading the same documents, businesses authorise access once. Trusted enterprise data becomes the collateral for cheaper micro, small, and medium enterprise credit, just as the Account Aggregator framework unlocked retail lending. While Udyam, GSTIN, and PAN are available via EntityLocker today, Ministry of Corporate Affairs documents and countless state-level entity licences and certificates are conspicuously absent. 

From PDFs to APIs: Over 2,500 government websites still require humans — not machines — to navigate them, each with its own interface and filing process without open APIs. Instead of people navigating portals, software — and increasingly AI agents — can communicate directly with government systems. Open APIs published through API Setu for all workflows, filings, and payments will embed government into business workflows. India already has 8,700 government APIs; the challenge is ensuring every compliance journey exposes one and reuses enterprise DPI. The alternative is AI agents resorting to password sharing and screen scraping, or anti-bot Captchas taking us back to square one. 

India Code: A live, single source of truth for all laws does not exist; obligations are scattered across three instruments (Acts, rules, and 25+ edicts) and central ministries/state governments, making discovery as difficult as compliance. The current physical gazette and e-gazette must be merged with the India Code to become a single, authoritative source of regulatory truth: A machine-readable repository of obligations that allows enterprise systems — and increasingly AI agents — to discover, interpret, and act on them digitally and autonomously. AI changes the economics of regulatory clarity and aggregation; digitising laws and notifications in machine-readable form in any state- or institution-specific portal enables AI to discover, structure and publish them through India Code. The challenge is no longer harmonising systems but digitising rulebooks. 

The biggest upside of enterprise DPI is a new imagination that accelerates and amplifies the new tone from the top about Jan Vishwas (citizen trust). A paperless, presenceless, and cashless interface for enterprises without deregulation and civil service reform has been criticised as paving the cowpath rather than building a highway. These critics have never been employers. 

The writers are, respectively, the co-founder of TeamLease Services and the co-founder of the Centre for DPI at IIIT Bangalore