India knows within minutes when a new company is born or a salaried worker is paid. It has had a permanent number for every taxpayer since 1972, has tracked tax deducted at source through employers and banks for two decades, and since 2021 shows each taxpayer a statement of what banks, registrars and employers have reported about them. It began giving patients a comparable number only in 2021, and joining is still voluntary for patients, doctors, hospitals and insurers. Over the same years, India’s IT companies built the tax system and a business processing health data for American and European insurers, hospitals and drug companies that brings in billions of dollars a year. The gap at home is not a shortage of technology or engineers. It is the difference between a system the state needed to collect money and a system it has chosen to fund modestly and leave optional.

How much to trust it. Company revenue shares come from filings and results releases as summarised by the press. Tax and health milestones come from Parliament replies, cabinet summaries and press reports; several government websites could not be opened, so some official figures are as relayed; the ABDM backgrounder, the Cabinet approval of the mission’s outlay, the PAN 2.0 explainer and Cognizant’s 10-K were read directly. Two figures rest on a single source and are marked: the count of health-sector capability-centre staff and the claim that only a third of the health mission’s money was spent. Where this piece does its own arithmetic, it says so.

Two identities, half a century apart

A clarification on dates. PAN is often said to have started in 2003. What happened in 2003 was that issuance was handed to UTI and NSDL and a new photo card was introduced, followed by NSDL’s Tax Information Network. The number itself dates from 1972, and the computerised ten-character format from 1995. The health side has no equivalent history before 2018: PM-JAY gave the state a claims system for one insurance scheme, and the digital mission that followed in 2021 is the first attempt at a patient identity for everyone.

What each system can do today

The difference is in the third and fourth rows. The tax system does not wait for anyone to opt in. Employers deduct tax and report it, banks report large transactions, every company is numbered at birth and every business above the GST threshold files electronically. The 80 crore PANs are not all active taxpayers, since only about 7.8 crore returns are filed a year, but the state can see the transactions behind them. The health system has the identity numbers, about 97.6 crore ABHAs, but the flow of records depends on facilities choosing to connect. Of 5.47 lakh facilities on the registry, about 1.3 lakh are counted as connected, and only about 17,000 of those are private, in a country where most outpatient care is private. The claims exchange that would carry insurance claims between hospitals and insurers had 34 insurers and administrators live in mid-2024; no later count or claims volume has been published, and as of August 2026 the insurance regulator was discussing incentives rather than a deadline.

One part of the health stack does run at tax-system scale: eSanjeevani, the government teleconsultation service built by C-DAC, passed 50 crore consultations in September 2026. It works because the government is both the provider and the payer, which is also why the tax system works.

Meanwhile, the export business

The same Indian IT industry has built a large business on other countries’ health systems. Cognizant, listed in the United States but with about 73% of its staff in India (256,900 of 351,600 at the end of 2025), earned US$6.35 billion from health sciences in 2025, about 30% of its revenue, much of it running claims and administration platforms for American insurers. TCS earned 10.4% of its FY2025-26 revenue from life sciences and healthcare, about US$3.1 billion by this blog’s arithmetic, or roughly Rs 27,000 crore at about Rs 87 to the dollar. That one year’s revenue is about seventeen times the digital health mission’s five-year outlay.

The work is claims adjudication, revenue-cycle management and medical coding for American hospitals, electronic-record implementation, and clinical-trial and drug-safety data for pharmaceutical companies. Infosys agreed in March 2026 to buy Optimum Healthcare IT, an American provider of hospital IT services, for up to US$465 million, and Cognizant was reported in July 2026 to have won a multi-year deal from the US insurer Centene worth more than US$500 million.

Foreign companies also run their own centres in India. An EY survey in September 2025 found that 23 of the world’s 50 largest life-sciences companies have a capability centre in India, most set up in the past five years. Novartis has more than 9,000 staff in India, about 11% of its global workforce; Sanofi has raised its Hyderabad centre’s target to about 4,500; Elevance Health’s Indian arm, Carelon Global Solutions, is estimated at about 22,900 people; and Walgreens signed up in September 2026 to open its first Indian centre in Chennai. One industry estimate, from a single secondary source, puts health and life-sciences capability-centre employment at about 3 lakh. No source found measures what share of any American insurer’s claims is processed from India, so this piece does not claim that India “runs” anyone’s health system; it handles a large and growing part of the back office.

Same engineers, different mandate

Three differences explain most of the gap.

1. Compulsion. The Income Tax Act made PAN compulsory for taxpayers; TDS and GST registration made employers, banks and businesses report on their own. No law requires a hospital, doctor or insurer to use ABHA, the professional registry or the claims exchange.
2. Money. Infosys’s three tax-system contracts alone come to about Rs 6,700 crore by this blog’s addition, before PAN 2.0. The digital health mission was approved at Rs 1,600 crore for five years, one report says only about a third of it was spent, and its 2026-27 budget is Rs 350 crore, 0.33% of the health ministry’s. Its incentive scheme for hospitals and labs had paid out about Rs 136 crore by June 2026.
3. Who pays whom. The tax system serves the state’s revenue, so the state paid for it and made it compulsory. In health the state is a minority payer, most care is private and paid out of pocket or by private insurers, and the parties who would have to record and share data gain little from doing so unless paid or required to.

The United States, whose hospitals Indian firms now serve, went the incentive-and-mandate route: after the 2009 HITECH Act tied payments to certified electronic records, 93% of its hospitals had one by 2013 and more than 99% by 2024. India’s incentive scheme is a small fraction of that effort.

How the rest of Asia does it

India’s neighbours took one of two shortcuts that India has not. Most use the national identity number as the health identifier instead of issuing a separate one, and most route claims through a single payer or clearing house, so that going electronic is the condition for getting paid. Several also set a legal deadline.

Three contrasts stand out. First, the mandates are dated. Japan required clinics and pharmacies to check insurance eligibility online from April 2023 and stopped issuing paper insurance cards in December 2024; Indonesia set 31 December 2023 for electronic records at every facility; Singapore, which ran its national record on a largely voluntary basis for private providers for about fifteen years, passed a law in January 2026 making contribution compulsory from early 2027. India has no such date for ABHA linkage or the claims exchange. Second, the identifier already exists. Korea, Thailand, Japan and Vietnam use the national ID; India issued a separate, voluntary health ID on top of Aadhaar, which is why a 97.6 crore account count can sit beside a quarter of facilities connected. Third, a single payer does most of the work. In Korea and Taiwan every claim passes through one national system, and in Taiwan paper claims are under 1%; India’s health spending is split between the government schemes, private insurers and patients’ own pockets, so no single payer can impose a format.

A mandate is not the whole answer. Japan’s My Number card was used in only about 37% of eligibility checks in October 2025, nearly a year after paper cards stopped, and Indonesian clinics were still preparing for their electronic-records deadline in 2024, with hundreds of separate health apps in use. But every country in the table that reports high connection rates has either a single payer or a dated legal requirement, and India has neither for health. It has both for tax.

The data-protection gap

India’s Digital Personal Data Protection Rules were notified in November 2025, with most obligations phasing in by about May 2027. The law has no separate category for health data; the 2018 draft health-data bill, DISHA, was not enacted. Section 17(1)(d) of the Act also exempts most obligations for the personal data of people outside India processed in India under a contract with a foreign company, which covers much of the export work above: American patients’ records processed in Indian centres are governed mainly by American law and contracts. Indian patients’ records, once the rules apply, will be covered by the general law alone.

What is not known

1. Volumes. No published figure for claims processed through the National Health Claims Exchange, or for how many facilities actually share records rather than merely register.
2. Spend. The “only a third spent” figure for the health mission is from one report; official expenditure for 2021-26 was not found.
3. Export scale. No industry-wide figure for healthcare’s share of India’s IT and business-services exports, and no measure of how many foreign patient records or claims are handled from India.
4. Builders. No evidence was found that TCS, Infosys, Wipro or HCLTech built the core of the health mission or the claims exchange; that is absence of evidence from a limited search, not proof.

What would fix this

• The insurance regulator should set a date by which every health insurer and every PM-JAY-empanelled hospital must exchange claims through the National Health Claims Exchange, as GST made e-invoicing compulsory above a turnover threshold and then lowered it in steps, and as Japan, Indonesia and Singapore set dates for their health systems.
• The National Health Authority should publish monthly volumes: claims through the exchange, records actually shared between facilities, and connected private facilities, the way GSTN publishes returns filed.
• The Centre should fund the health stack at tax-system scale and buy it the same way, through long-term managed-service contracts with service-level penalties, as it did for the Centralised Processing Centre and GSTN.
• Pay hospitals for records shared, not for registration, and stop routing incentive money to software vendors.
• Give health data its own protections under the data-protection rules before linkage becomes compulsory.

These are proposals from this blog, not recommendations any agency has adopted.

Conclusion: the jobs AI takes abroad, and the ones India could make at home

The export business described above is the part of this story most exposed to AI. Medical transcription is the clearest case. The US Bureau of Labor Statistics projects employment of medical transcriptionists to fall about 5% between 2024 and 2034, from roughly 43,900 to 41,800, and ambient AI scribes are removing the step that sent work offshore: the note is now written in the consulting room while the doctor talks. Abridge says its scribe is live in more than 300 health systems, and Epic, whose software runs most large American hospitals, announced its own AI charting tool in August 2025 for use from early 2026 (Digital Health). Coding and revenue-cycle work come next. Indian firms describe AI as raising output per worker rather than cutting staff so far: Omega Healthcare’s chief executive reports 10–20% higher productivity and expects at least 20% more (Healthcare IT News), and GeBBS says autonomous coding lets volume grow without adding headcount. No layoffs attributed to AI were found at these firms, and no credible estimate exists of how many of India’s health back-office jobs, put at about 3 lakh by one consultancy, are at risk. The first effect is likely to be fewer new hires for each dollar of revenue rather than sudden cuts.

India’s own system has the workload that could absorb those skills. PM-JAY authorised 2.74 crore hospital admissions in 2025-26, and private insurers processed 3.26 crore health claims in 2024-25, of which 8% were rejected (Business Standard on IRDAI data). That is roughly 6 crore claims a year by this blog’s rough sum, across different years and with some overlap, each needing pre-authorisation, coding, documentation and audit: the same work India sells to American hospitals. About three-quarters of the 5.47 lakh registered facilities are not yet connected to the digital health mission, and connecting each one means setting up systems, digitising records and running ABHA and Scan-and-Share at the counter. India’s languages add work AI cannot yet do alone. A February 2026 agreement between Bhashini and the National Health Authority names clinical documentation as a use for Indian-language voice AI (Digital Health News), and startups such as Eka Care offer Indian-language scribes, but research benchmarks published in 2026 show Indian-language medical speech remains an unsolved problem. Building and checking those models needs people who know Indian languages and Indian clinical practice, in India.

Whether that becomes employment depends on the same two things that built the tax system: a mandate and money. A dated requirement to use the claims exchange would create demand for coders and claims staff at hospitals and insurers; paying facilities for records actually shared would pay for digitisation; and a funded programme to build clinical speech and text data in Indian languages would employ annotators and clinical reviewers. None of these has a published job estimate, and the only official figure in this area, the Prime Minister’s 2019 statement that PM-JAY would create 11 lakh jobs, came with no stated method. The difference between the two markets is the point. In America, AI is removing routine work from a system that is already fully digital. In India, the system is not yet digital, and the same tools would first need people to build the record. The tax system shows India can build the pipes when it decides to. For health, that decision is still pending, and many of the people who could build it are, for now, writing other countries’ medical notes.

Corrected 25 October 2026. Cognizant’s share of staff in India is about 73% (256,900 of about 351,600 employees at 31 December 2025, per its 10-K), not “more than 85%”; the revenue figures (US$6.35bn, 30%) are unchanged. The text, stat card and chart label are corrected.

Sources and caveats

About this article: Researched, written and edited by Umashankar Triplicane Dwarakanathan, with AI research assistance; every figure is meant to trace to the primary source cited. See the Editorial Policy for how sourcing, AI use and corrections work.