How AI Is Transforming India’s Key Sectors
India’s AI adoption is ahead of schedule. Its accountability architecture is still catching up.
Topics
News
- Nvidia Enters the PC Market with RTX Spark to Run AI Agents Locally
- OpenAI Launches Biodefense Program for Public Health Agencies
- YouTube Moves AI Labels to More Prominent spots
- TCS Partners With Mistral to Build Enterprise AI Models
- Pronto Faces Scrutiny Over In-Home Camera Pilot
- AI Dispatch | 22 - 29 May
India is emerging as a global leader in AI talent and adoption. The country ranks first in AI skill penetration in Stanford’s 2026 AI Index, and Indian professionals are among the world’s most active users of AI tools at work, with more than 80% reporting regular use. At the same time, enterprises across sectors are moving beyond experimentation toward scaled deployments. Government programs have embedded AI into tuberculosis treatment, weather forecasting, judicial translation, grid management, and battlefield targeting.
By almost every measure of adoption, India is ahead of schedule; 85 to 90% of organizations report actively backing AI strategy and governance, the highest share of any country surveyed. What is harder to measure is whether that deployment is producing the outcomes it was designed for, at the populations it was meant to serve, with safeguards sufficient for the consequences involved. No clean answer exists yet in any of the ten sectors examined below.
For large Indian enterprises, the shift is also moving beyond automation into operating-model redesign. Vijay Balakrishnan, Chief Digital and Information Officer at Godrej Enterprises Group, says India’s AI boom is being shaped as much by real-world constraints as by technological ambition.
“India’s technology boom is often framed around automation and efficiency. With AI, we are seeing a deeper shift that’s impacting all industries,” Balakrishnan says.
He says companies are beginning to move beyond platform-level AI features toward systems that connect multiple business processes and deliver measurable stakeholder value, from traceable factory operations to connected products that feed customer insights back into design and R&D.
The MIT Sloan Management Review India AI Research Forum convenes in Bengaluru on 23 July at a specific moment in that story. The pilots are over and the systems are live. India is now working out whether the governance, accountability, and equity architecture surrounding those systems can keep pace with their scale.
That architecture is lagging in two distinct ways, not one. In six of the sectors below, the gap is one of accountability: when an AI system errs, no statute fixes who is liable, and the affected person has no route to challenge the output. In three more, the gap is one of distribution. The benefits and costs are unevenly distributed, leaving the base of the pyramid (MSMEs, informal workers, non-English speakers, and remote farmers) either unreached by the tools or absorbing their downsides without representation. Education sits between the two, deciding whether India can train enough people to close the first gap and widen access against the second. Seen through those two failures, the ten sectors no longer look alike.
1. HEALTHCARE
The clinical numbers are the strongest in any sector. The legal accountability is the weakest.
A government statement published on 13 February set out what three years of AI integration into India’s public health system had produced. AI-enabled tools within the National Tuberculosis Elimination Program (NTEP) contributed to a 27% decline in adverse TB outcomes. The DeepCXR system automates chest X-ray analysis to detect presumptive TB across eight states, reducing dependence on radiologists that the public system does not have. MadhuNetrAI has screened 7,100 patients across 38 facilities for diabetic retinopathy using portable devices that non-specialists can operate. The Media Disease Surveillance System has issued more than 4,500 outbreak alerts since April 2022. The e-Sanjeevani telemedicine platform logged 282 million consultations between April 2023 and November 2025, with AI-assisted diagnoses helping 12 million patients.
The shift Radha Basu, Founder and Chief Executive of data annotation and AI services firm iMerit, describes goes beyond the headline numbers. Generic AI models trained on global datasets are giving way to systems built by clinicians, cardiologists familiar with Indian cardiac presentations, radiologists who know TB profiles in high-burden districts, and pathologists experienced with edge cases a model trained elsewhere would miss.
“The key challenges ahead are trust, explainability, and data quality. Healthcare AI cannot rely solely on internet-scale data. Models must be continuously tested, challenged, and refined by experts who can identify hallucinations and edge cases before deployment.”
Radha Basu, Founder and Chief Executive, iMerit
REALITY CHECK
The clinical numbers are real. The legal framework around them is not finished. No statute assigns responsibility when an algorithm contributes to a missed diagnosis. No mechanism requires that a patient be told an AI system was involved in their care, or gives them the right to question its output. The National AI Doctors Mission, launched last month, and the IndiaAI-ICMR memorandum of understanding represent a serious institutional response. Neither is enforceable law.
2. AGRICULTURE
Better data is reaching farmers. The last mile remains the hardest.
Agriculture employs nearly 46% of India’s workforce and contributes 18% of gross domestic product (GDP). The tools being deployed are substantive. The government’s Kisan e-Mitra virtual assistant helps farmers navigate scheme applications including the PM Kisan Samman Nidhi income support program. The National Pest Surveillance System combines satellite imagery, weather data, and soil analysis to provide real-time crop advisories. This year, the India Meteorological Department (IMD) launched an AI-enabled monsoon advance-forecasting system producing probabilistic outlooks up to four weeks ahead. Most significantly, the Bharat Forecasting System (BharatFS), launched on 26 May 2025 by the Ministry of Earth Sciences, delivers weather predictions at 6 km resolution, down from the previous 12 km models, with rainfall forecasts accurate up to 10 days ahead. It is currently the only operational global weather prediction model running at that resolution in real time. Sankara Subramanian, Managing Director and Chief Executive of agriculture and crop solutions firm Coromandel International, describing the company’s platform at an industry event, said: “In agriculture, AI is accelerating the shift toward precision farming through personalized and data-led solutions. Our digital platform MyGromor, which today serves over one million users, leverages AI-enabled capabilities such as crop advisory, pest and disease detection, and plant health diagnostics. We are also integrating vernacular AI chatbots to make advisory services more accessible to farmers across regions.”
Basu adds the structural qualifier. “Agritech AI in India is earlier in its maturity curve, but has enormous long-term potential,” she says. India’s diversity of crops, climates, languages, and farming practices makes localization critical. Global foundation models only become useful when agronomists, farmers, and regional experts embed crop, soil, weather, and seasonal intelligence into the systems.” She adds that multi-sensor satellite imagery is also improving the underlying understanding of land and water that AI advisory systems draw from.
REALITY CHECK
The data infrastructure for agriculture is improving faster than the localization of AI systems that run on it. BharatFS delivers village-level weather data that did not previously exist. MyGromor reaches a million users. What does not yet exist at scale is the expert feedback loop: agronomists, soil scientists, and farmers embedded in the model-building process across India’s 700 distinct agro-climatic zones, which is needed to make systems reliable for planting or treatment decisions.
3. FINANCE
Financial inclusion is expanding. The bias question is not resolved.
India’s financial services sector is built on a digital infrastructure that most countries are still trying to replicate: Aadhaar biometric identity, the Unified Payments Interface (UPI), the Account Aggregator framework, and now the Unified Lending Interface (ULI). AI-driven credit scoring on top of this stack is reaching people who have never been formally assessed before. The Reserve Bank of India’s (RBI’s) Mulehunter AI tool, launched in December 2024, analyzes transaction patterns in real time to detect mule bank accounts used in cybercrime. In February, the Digital India BHASHINI Division and the RBI signed a memorandum to integrate multilingual AI into banking services across all 22 scheduled Indian languages.
Aditya Agarwal, co-founder of Wealthy. in, a wealth management platform for mutual fund distributors, describes what that means at the individual level. “As millions of first-time investors enter the market, particularly from Tier-2 and Tier-3 India, many are navigating investing for the very first time,” he says. “The confidence to invest meaningfully, understand risk, and stay disciplined through market volatility is still developing. This is where AI is beginning to make a meaningful difference.”
Pragati Awasthi, an assistant professor of AI and Data Science at US-based Drexel University, frames the sector’s central tension. “AI-driven credit scoring is now reaching populations being formally assessed for the first time, and this is a genuine opportunity and a meaningful development for financial inclusion,” she says. “But every market is working through whether explainability and bias controls are being built alongside deployment, or after the fact. The RBI’s FREE-AI framework reflects that honest tension. The infrastructure for accountability is still catching up to the scale of adoption.”
The RBI’s Framework for Responsible and Ethical Enablement of AI (FREE-AI), released in August 2025, documented its own baseline: only 20% of financial institutions have adopted AI beyond basic rule-based scenarios, and fewer than 15% conduct post-deployment monitoring for bias or performance drift.
REALITY CHECK
The inclusion story is one of the more consequential things happening in Indian finance. The same systems extending credit to previously unserved populations can also encode historical exclusion if their training data reflects past patterns and their outputs face no challenge mechanism. FREE-AI names this problem. The distance between naming it and addressing it systematically across thousands of lending institutions is where the next phase of this story will play out. Globally, the trend runs the other way: Stanford’s 2026 AI Index found developer disclosure actually fell last year, with its transparency index dropping from 58 to 40 and the widest gaps in training data and post-deployment monitoring, the two things a regulator would most need to see. India’s own baseline sits below even that, with fewer than 15% of financial institutions monitoring for bias or drift after deployment.
4. LOGISTICS AND SUPPLY CHAIN
Measurable efficiency. No one in charge when it goes wrong.
Saurav Swaroop, Co-Founder and Chief Technology Officer of Velocity, an e-commerce enablement firm, describes what the operational shift looks like from inside a logistics business. “AI is now fundamentally transforming this landscape by automating large parts of these workflows at scale,” he says. “One of the most impactful use cases is shipment intelligence and delivery operations.”
AI systems identify and classify non-delivery attempts, distinguish genuine failures from fraudulent ones, and flag high-risk shipments before they enter the fulfillment pipeline.
Swaroop describes one specific deployment at Velocity: “We have been actively deploying AI-driven automation across multiple parts of the logistics workflow. One example is the deployment of AI voice agents for order confirmations and for verifying non-delivery reports. If a customer requests delivery on a different date during a call, that information is automatically captured and relayed to the logistics provider in real time. Traditionally, achieving this level of coordination would have required setting up large call center operations. AI enables the same, often with greater speed, consistency, and scalability.”
That shift is beginning to move beyond workflow automation. Ed Huang, Co-founder and Chief Technology Officer of TiDB, says India’s retail, logistics and e-commerce sectors are moving from basic AI assistants toward more autonomous operational systems.
“India’s market is highly cost-sensitive, fragmented, multilingual, and mobile-first, which makes automation especially valuable,” Huang says. “One important shift is that AI systems are becoming increasingly agentic and autonomous.”
Huang says this will create a new layer of infrastructure requirements. Future systems will need to support large numbers of AI agents coordinating with one another, accessing shared memory and interacting with enterprise data in real time.
Speaking at the India AI Impact Summit, Gourav Vallabh, a member of the Economic Advisory Council to the Prime Minister, estimated that AI could save nearly ₹20,000 crore in cargo handling and another ₹15,000 crore annually in broader port logistics if deployed consistently across major ports and corridors. A KPMG report on AI in Indian supply chains found that disruptions cost Indian companies close to $12 billion annually, giving proactive AI-driven logistics management a clear financial case.
REALITY CHECK
The efficiency gains are documented and the numbers are large. The gap is in what happens when the system fails. When an AI-managed logistics prediction cascades into a supply chain failure, no published framework in India specifies who carries liability, what the affected party’s recourse is, or which regulator owns the outcome. The sector is building faster than the legal architecture around it.
5. EDUCATION AND SKILLING
The hinge sector. Strong reach, but a curriculum that trains people to build systems rather than question them.
India’s AI education footprint is genuinely wide. The Central Board of Secondary Education (CBSE) now offers a 15-hour AI skills module from Class VI and an optional AI subject from Class IX through Class XII under the National Education Policy (NEP) 2020. The DIKSHA platform deploys AI for accessibility features. The YUVAi program equips students from Classes VIII to XII with AI problem-solving skills across eight thematic areas. The India Skills Report 2026, produced by ETS with the Confederation of Indian Industry (CII), the All India Council for Technical Education (AICTE), and Taggd, found that more than 90% of Indian employees have already begun using generative AI tools. The Stanford AI Index 2026 ranks India second in the world for total AI authors and inventors, behind only the US.
Pragati Awasthi, who works on AI curriculum design across institutions, identifies the gap those numbers conceal. “India is producing technically strong graduates,” she says. “What the EY-FICCI 2025 report on higher education surfaces, and what I see in curriculum design conversations, is that most programs are not yet teaching students to critically interrogate the systems they build. Knowing how to train a model and knowing when to question it are different competencies. We need both.”
REALITY CHECK
India is producing AI-capable engineers faster than in almost any other country. What the curriculum does not yet teach at scale is how to evaluate a system’s assumptions, its failure modes, and whether it should be deployed in a particular context. This matters most in the sectors where AI is now making decisions that affect people who cannot read the model’s output or know it was involved.
6.MANUFACTURING
Globally competitive at the top. The MSME gap is widening.
India’s smart factory market is valued at $7.7 billion in 2025 and is projected to reach $17 billion by 2032 at a 12% compound annual growth rate, according to research from P&S Intelligence. The Production Linked Incentive (PLI) scheme has generated ₹1.46 trillion in investments across 14 sectors as of August 2025, creating about 950,000 jobs. Tata Steel’s Kalinganagar plant holds a World Economic Forum (WEF) Global Lighthouse designation for its Industry 4.0 integration. Across automotive, electronics, and precision engineering clusters in Pune, Chennai, and Surat, predictive maintenance, AI-powered quality inspection, and energy management systems are live at scale.
The National Mission on Interdisciplinary Cyber-Physical Systems (NM-ICPS) under the Department of Science and Technology is supporting AI integration in manufacturing through 25 technology innovation hubs nationwide. Industrial robot installations in India reached a record 8,510 units in 2023, with the operational stock growing to 44,958 units. The numbers reflect the pace of automation at the leading edge of the sector.
The constraint is structural and concentrated at the base. A March 2026 industry analysis found that a basic predictive maintenance pilot on three to five machines costs ₹10 to 15 lakh, and full factory digitization runs ₹50 lakh to ₹2 crore for a mid-size facility. Most Indian factories run machines that are 10 to 20 years old, built before connectivity was a design consideration. For micro, small and medium enterprises (MSMEs) operating on thin margins, the math rarely works even when the two-year return on investment case is clear.
REALITY CHECK
The top of India’s manufacturing sector is among the most AI-capable in the world. The base, which consists of hundreds of thousands of MSMEs that employ the majority of manufacturing workers, is largely unreachable by the same tools. The PLI schemes are accelerating the leading edge. The Industry 4.0 demonstration centers are beginning to address the middle. Nothing currently operating at scale is closing the gap at the base, and the competitive pressure from AI-capable manufacturers is already being felt by those who have not made the transition.
7. DEFENSE AND NATIONAL SECURITY
Combat-proven capability. The governing doctrine is still being written.
Operation Sindoor last May produced the most specific public account of AI in live military operations anywhere in the world that year. Speaking in October 2025, Lt. Gen. Rajiv Kumar Sahni, who served as Director General of Information Systems during the operation, said 23 AI applications were integrated across the operation. By processing 26 years of historical radio emission and frequency signature data, the system achieved more than 94% accuracy in locating and targeting Pakistani military assets. Military AI decision-support system Anuman 2.0 delivered real-time weather forecasts 48 hours ahead with 200 km border precision, feeding directly into artillery calculations. “AI was extensively used for multi-sensor and multi-source data fusion in real time during Operation Sindoor,” Lt. Gen. Sahni said.
The institutional architecture around this capability has been building steadily. The Centre for Artificial Intelligence and Robotics (CAIR), a Defence Research and Development Organisation (DRDO) laboratory, has developed autonomous vehicles, AI-enabled surveillance tools, and military-grade speech recognition systems. The Indian Army designated 2024-25 the Year of Technology Absorption and has established dedicated AI cells across commands. A military-specific large language model (LLM), Lt. Gen. Sahni indicated, was expected to be operational within six months of October 2025.
REALITY CHECK India has confirmed AI capability in combat conditions and can document what it did. The doctrinal gap is significant. Using 23 AI applications in an operation is one thing. Having a published framework that governs how those applications are authorized, how their outputs are validated before a targeting decision is made, and who bears accountability when they are wrong is another. The technology absorbed faster than the doctrine written to govern it. |
8. GOVERNANCE AND JUDICIARY
Specific deployments. The citizen on the receiving end has no recourse yet.
India’s judicial AI deployment is among the best-documented in the world. The Supreme Court Vidhik Anuvaad Software (SUVAS) has translated over 36,271 Supreme Court judgments into Hindi and 17,142 into 16 regional languages, according to a Frontiers in Political Science study published in 2025. The Supreme Court Portal for Assistance in Court Efficiency (SUPACE) helps judges identify precedents. In November 2025, the Kerala High Court mandated that all subordinate courts use Adalat, an AI speech-to-text tool, for recording witness depositions, a system-wide rollout affecting thousands of courts across an entire state.
This February, the Supreme Court issued comprehensive AI guidelines for judicial administration: AI permitted for case listing, legal research, translation, and docket management; AI explicitly prohibited from replacing judicial reasoning. The e-Courts Phase III project has allocated ₹7,210 crore for judicial digitalization, with ₹53.57 crore earmarked for AI and blockchain integration across High Courts.
Awasthi names the central issue. “India’s AI moment is the gap between the deployment curve and the governance curve,” she says. “The question is how quickly that gap could become consequential.” India’s November 2025 AI Governance Guidelines from the Ministry of Electronics and Information Technology (MeitY) establish a principle-based framework anchored to existing statutes. They do not establish enforcement.
REALITY CHECK The judiciary’s AI work is specific and consequential. Fifty-four million pending cases represent a governance failure that translation and research tools can genuinely address. What the current framework does not answer is what happens when an AI system deployed by the state makes a material error affecting a citizen who cannot read English and has no way of knowing an algorithm was involved in their case. And the risk is not evenly spread: Stanford’s 2026 AI Index found these systems work best in English and degrade sharply in regional languages, with several leading models losing close to half their accuracy on a regional dialect. Across 22 scheduled languages, the citizen least able to catch the error is the one the system is most likely to make it on. That gap needs a statutory answer before deployment grows further.
|
9. ENERGY AND CLIMATE
The grid needs AI to function. The rules for AI-run infrastructure don’t exist yet.
India’s energy sector faces a specific AI constraint that is not about efficiency. With renewable energy projected to exceed 500 gigawatts (GW) before 2030, grid variability is approaching the point where human-speed decision-making is insufficient to maintain stability. Samir Chandra Saxena, Chairman and Managing Director of national electricity grid operator Grid India, told a KPMG energy conference in September 2025 that India is planning to deploy AI into the national grid using real-time streaming data at 40-millisecond resolution to detect instability before it cascades into outages. “The grid’s physical and digital layers are now inseparable,” he said. “As variability increases, we can’t depend on post-fault analysis alone. AI must give us situational awareness before faults propagate.”
On the climate side, the Bharat Forecasting System (BharatFS), launched on 26 May 2025, operates at 6 km resolution and delivers 10-day rainfall forecasts. It is 30% more accurate at predicting extreme rainfall than its predecessor. The Central Water Commission uses AI for short-range flood forecasting. The National Disaster Management Authority (NDMA) deploys AI-based tools for cyclone preparedness and evacuation modeling. India’s data center electricity demand is projected to reach 13.56 GW by 2031-32, up from a base of 1,500 megawatts (MW) of total data center capacity in 2025.
REALITY CHECK
The energy sector is deploying AI because the alternative is an unstable grid. That urgency is different from any other sector in this feature. What India does not yet have is a published governance framework for what happens when AI-managed power infrastructure fails. Automated systems operating at 40-millisecond resolution cannot be paused for a committee meeting. The accountability architecture for machine-speed grid management needs to be designed before the grid reaches the scale at which a systemic failure would be nationally consequential.
10. WORKFORCE AND THE FUTURE OF WORK
Strong at the top of the skill curve. The informal base has no protection.
The India Skills Report 2026, produced by ETS with CII, AICTE, and Taggd, found that more than 90% of Indian employees have already begun using generative AI tools. National employability has risen from 46.2% in 2022 to 56.35% in 2026. India accounts for 16% of the global AI talent pool, projected to reach 1.25 million professionals by 2027. The gig workforce grew by 55%, from 7.7 million in FY21 to 12 million in FY25, and is projected to reach 23.5 million by 2029-30, according to a Press Information Bureau note.
That is the top half of the distribution. An Observer Research Foundation analysis published in April 2026 found that over 90% of India’s workers are employed informally, with the e-Shram Survey finding over 150 million workers in the informal non-agricultural sector alone. The same analysis identified three specific AI displacement pathways for this group: micro-retailers and street vendors being undercut by AI-enabled e-commerce platforms; platform gig workers having their bargaining power eroded by algorithmic management; and informal manufacturing tasks being automated without any retraining pathway available. India’s Economic Survey 2025-26, tabled in Parliament on 29 January, acknowledged the risk directly: gig workers are treated as a largely homogeneous group in law, even though their actual skill and vulnerability profiles vary widely.
Awasthi connects the workforce and governance threads directly. “The gap between deployment enthusiasm and measurable business value is becoming almost harder to ignore,” she says. “This is not because the technology is weak. Organizations change much more slowly than technology does.” For informal workers, those organizations are not changing at all.
REALITY CHECK India’s AI workforce story is being measured from the top: talent pools, employability rates, and gig economy growth. The more consequential story sits at the bottom: 150 million-plus informal non-agricultural workers absorbing the economic effects of AI deployment in sectors where they have no representation, through displacement pathways that no current policy identifies as a priority. The Economic Survey acknowledged the risk in January 2026, but acknowledgment and architecture are not the same thing. And the gap is structural, not just rhetorical: the World Economic Forum’s Future of Jobs Report 2025, the most authoritative global survey of its kind, draws on a dataset of 1.2 billion formal jobs, which means it cannot see the roughly 90% of India’s workforce that is informal. The population most exposed to AI displacement is the one the instruments are not built to count. |
Where Ten Sectors Converge
Across the ten sectors the same thing is being built faster than the framework meant to govern it. But the framework is failing in two different ways, and the distinction is the point. In six sectors the failure is one of accountability. In healthcare the clinical results are documented but the liability law is not written; in defense the targeting accuracy is confirmed but the authorization doctrine is not published; in finance the credit access is expanding but the bias monitoring is minimal; in logistics the efficiency is measurable but the failure accountability is undefined; in the judiciary the deployment is specific but the citizen on the receiving end has no recourse; in energy the grid is preparing for machine-speed management with no published framework for machine-speed failure.
In each, the question is the same: when the system is wrong, who answers for it? In three further sectors the failure is one of distribution. In agriculture the data reaches the village but the localization does not; in manufacturing the leading edge is world-class but the MSME base cannot afford the entry ticket; in the workforce the talent pool is measured while the informal majority absorbs the displacement unrepresented and uncounted. There the question is not who is liable but who is left out. Education sits between the two, the one sector whose progress determines whether either gap can be closed, because closing them needs people trained not just to build these systems but to interrogate them.
This is not unique to India. The accountability lag is global: the AI Incident Database logged 362 documented incidents in 2025, up from 233 a year earlier, even as model developers report capability scores routinely and responsible-AI results barely at all, per Stanford’s 2026 Index. What is distinctive about India is the scale at which it is running the experiment and the speed at which the systems have moved from concept to national deployment. No other country has embedded AI into tuberculosis treatment, judicial translation, battlefield targeting, and agricultural forecasting in the same decade. In some cases, the same year.
ABOUT THIS FEATURE Primary interviews with Radha Basu (iMerit), Pragati Awasthi (Drexel University), Aditya Agarwal (Wealthy.in), Saurav Swaroop (Velocity), and S. Sankarasubramanian (Coromandel International, speaking at an industry event). Policy and data sources: Government of India PIB healthcare AI release (February 2026); BharatFS launch documentation, Ministry of Earth Sciences (May 2025); Grid India/KPMG energy conference coverage, Down to Earth (September 2025); Tribune India on Operation Sindoor AI (October 2025); Frontiers in Political Science SUVAS study (2025); ORF India Informal Sector and AI (April 2026); Economic Survey of India 2025-26 (January 2026); India Skills Report 2026 (ETS/CII/AICTE/Taggd); P&S Intelligence India Smart Factory Market; RBI FREE-AI Framework (August 2025); India AI Governance Guidelines, MeitY (November 2025); Supreme Court AI Guidelines (February 2026); KPMG AI in Indian Supply Chains (2026). Additional survey data: Stanford HAI Artificial Intelligence Index Report 2026, Responsible AI and Public Opinion chapters (April 2026); World Economic Forum Future of Jobs Report 2025 (January 2025). |
