The cleanest single fact about AI in India in 2026 is not a productivity statistic. It is a layoff number. TCS confirmed roughly 12,000 mid- and senior-level role exits in FY26. Infosys has eliminated more than 8,000 roles. Wipro has cut 5,500. HCL has cut 4,200. Campus hiring across the top Indian IT services firms is down approximately 60 percent.

This is not what the AI economy was supposed to look like when the prediction was first written. The Indian IT pyramid was meant to absorb the bottom of the global services demand for another decade. Instead, in the space of about twenty-four months, the same firms have re-staffed their delivery models around generative AI — and the people most exposed are the ones who used to sit on the second and third rungs of the pyramid.

For anyone running a business in India today, the honest read is not the boardroom-deck version of "AI changes everything." It is calibrated, two-speed, and far more interesting.

What India is actually building

India does have a sovereign AI stack and the foundations of one are real. The IndiaAI Mission, approved by the Union Cabinet at ₹10,372 crore in March 2024, has by late 2025 onboarded more than 38,000 GPUs and approved 190 AI projects.

The headline names worth knowing:

Sarvam AI (Vivek Raghavan and Pratyush Kumar, ex-AI4Bharat) was selected under the IndiaAI Mission in April 2025 to build India's sovereign foundational large language model. In February 2026 they open-sourced their 30B and 105B Indic-language reasoning models — the first production-grade attempt at a multilingual LLM that handles Indian languages on the same footing as English.

Krutrim, the Ola Group's AI bet led by Bhavish Aggarwal, became India's first AI unicorn in January 2024 — and in May 2026, pivoted from building foundation models and AI chips to running an AI cloud-services business after running into the structural cost-and-talent gap. That pivot is the single most useful corrective to the simple "Indian foundation models are winning" narrative.

Yotta Data Services (Hiranandani Group) is the country's GPU backbone — first NVIDIA H100 shipment in March 2024, scaling to roughly 32,000 GPUs by end of 2025, and a $2 billion AI hub being built with NVIDIA.

AI4Bharat at IIT Madras, where Mitesh Khapra (TIME 100 AI 2025), Pratyush Kumar and Anoop Kunchukuttan have built much of the Indic-language data and models behind the government's BHASHINI programme — the substrate on which Sarvam and others now train.

Around them sit the older corporates that have, quietly, become AI-deployment shops. HDFC Bank runs a real-time fraud engine analysing more than a hundred parameters per transaction and has cut card-fraud losses 20 percent. ICICI Bank runs iPal — its natural-language banking chatbot — and an enterprise fraud-risk management engine across all retail channels. Zomato runs computer-vision-based menu digitisation across millions of restaurant pages. Swiggy runs a 50-million-item neural search LLM across the consumer app.

Microsoft's Work Trend Index 2025 puts India at the top of the global league table on AI Advantage at work: 93 percent of Indian leaders plan to deploy AI agents in the next 12-18 months, and 86 percent of Indian employees report that AI has had a positive impact on their productivity.

These two stories — the bottom-of-the-pyramid contraction and the top-of-the-pyramid amplification — are happening to the same workforce, in the same year, in the same sectors. That is what bifurcation actually looks like.

Where the compression hits hardest

NITI Aayog's October 2025 Roadmap for Job Creation in the AI Economy puts the directly-affected figure at roughly 2 million Indian tech-sector jobs, which sit on an ecosystem of another 20 to 30 million dependent roles. Worst-case projection: the IT services headcount falls from approximately 7.5 to 8 million in 2023 to about 6 million by 2031. Customer-experience headcount falls from 2 to 2.5 million to roughly 1.8 million.

The concrete entry-level roles most exposed: call-centre and customer support agents, junior software engineers writing maintenance code, basic data-entry and back-office processing, junior content writers, paralegal and clause-checking analysts, junior accountants doing reconciliation work, basic translation, junior graphic and presentation design.

These are precisely the kinds of jobs that India built an entire export industry around twenty years ago. They are also precisely the kinds of jobs that the current generation of LLMs and AI agents can do at a fraction of the cost.

"AI doesn't replace humans. It replaces humans who don't use AIand in India that hits the bottom of the IT pyramid first, the captain on the restaurant floor last."

Where the amplification happens

The mid-career professional who learns to use AI well is the structural winner of the next five years. EY's 2025 Work Reimagined survey puts the average productivity gain at 5.44 percent across the Indian workforce that uses AI — a number that on its own sounds dull, but compounds dramatically over a five-year horizon.

The senior software architect, the brand strategist, the M&A analyst, the design director, the senior accountant, the experienced lawyer — every one of these roles becomes more productive, more scoped and more leveraged when AI takes over the rote layer underneath. The Indian premium for these roles is not falling. It is, in 2026, rising sharply.

Where AI doesn't go

The most under-discussed implication of the AI economy is that it makes certain categories of work more valuable, not less. Hospitality service. Healthcare frontline. Skilled trades. Last-mile logistics. Care work. Senior sales relationships. Creative direction at the top level. Anything that requires a hand, a heart, and a body in the room.

You cannot prompt-engineer a warm welcome at a hotel front desk. You cannot AI a steady pour at a busy bar on a Saturday night. You cannot replicate the instinct of a captain reading a table that is about to ask for the bill. The jobs that depend on being physically present, socially attuned and operating in real time — these jobs, the ones the Indian hospitality industry has spent thirty years building — are not the casualties of the AI economy. They are its quiet beneficiaries.

That does not mean they are easy to fill. It means the people who do them well will be priced higher, retained more carefully, and trained more deliberately than they have been at any point in recent Indian labour history.

The Indian-language gap

There is one structural problem unique to India that none of the global AI narratives capture. India has 22 official languages. Not one production-grade LLM, in May 2026, speaks all of them with equal fluency. Sarvam's open-source releases are the first serious attempt to close the gap. The gap is wide, the closing will take years, and until it closes, the AI economy will keep speaking English at a country that mostly does not.

For brands operating in tier-2 and tier-3 India — QSR chains, retail, hospitality, financial services — this is a constraint and an opportunity in the same sentence. The first AI assistant that handles a customer-service call fluently in Marathi-Hindi-English code-switch is going to take an entire vertical of customer experience and rebuild it.

What an honest takeaway looks like

For an Indian operator, hospitality executive, or HR head reading this in 2026, the working playbook is roughly:

  • Adopt AI hard for back-of-house work. Demand forecasting, revenue management, partner support, fraud, payroll, recruitment top-of-funnel, content. The cost-benefit is short, the upside is real, the risk is low.
  • Protect and re-value front-of-house human work. Captains, bartenders, chefs, front office, housekeeping leadership. Train them better, pay them better, retain them harder. The jobs you build around will become more, not less, scarce.
  • Watch what your IT-services partner is doing carefully. A 60-percent drop in their campus hiring is not their problem alone — it changes the labour market your own technology team draws from.
  • Treat AI as the new electricity, not the new internet. The companies that benefit are not the ones that talk about it most. They are the ones that have integrated it into a real workflow that was already broken — and noticed quietly that it now works.
"Hospitality is the anti-fragile industry of the AI decade. You cannot prompt-engineer a warm welcome, a steady pour, or the instinct that the table in the corner is about to ask for the bill. The jobs that need a hand and a heart just got more valuable, not less."

India's AI economy is real, growing, and uneven. The doom version of the story is incomplete. So is the utopia version. The truthful version, in the middle, is the one operators should be paying attention to — because the labour market your business will be hiring from in 2028 is being reshaped right now.

Kaam Hire is the hospitality-only hiring platform that powers this blog. The kind of jobs we build the brigade around — captain, bartender, sommelier, front office, head chef — are the jobs the AI decade is going to make more valuable, not less.

Sources & references 8
  1. PIB — IndiaAI Mission ₹10,372 crore approval (March 2024)
  2. NITI Aayog — Roadmap for Job Creation in the AI Economy
  3. EY India — 2025 Work Reimagined: India leads on AI Advantage
  4. Microsoft Source Asia — India's workforce goes AI-first (2025)
  5. NASSCOM — GenAI Tracker H2 FY25
  6. Business Standard — Sarvam launches Indic LLMs (Feb 2026)
  7. TechCrunch — Krutrim pivots to cloud services (May 2026)
  8. SME Futures — TCS 12,000 layoffs & AI restructuring

Kaam Hire is the hospitality-only hiring platform that powers this blog. If hiring is on your mind — try it.