AI and Jobs in India (2026): The Numbers That Hold Up and the Ones That Don't

In one dataset, India has 2.6 hires linked to AI (Artificial Intelligence) for every layoff or exit linked to it. In one survey, 55% of IT (Information Technology) firms say entry-level employment fell. Both findings are real, and neither is a headcount of India's economy.

Is AI creating or destroying jobs in India in 2026? On the evidence checked on 4 November 2026: AI-linked hiring is ahead of AI-linked layoffs in Nomura's tally of company announcements (69 across Asia), more IT firms report falling entry-level employment than rising in the ICRIER (Indian Council for Research on International Economic Relations) survey of 651 respondents, and no source we found counts every job in India, so the net effect on the whole economy is unknown. This guide grades each widely quoted number as holds up, holds up with limits, or cannot be traced, and says what each one measures and what it cannot show.

Last checked 4 November 2026. Every figure below was checked against the original document wherever it could be opened, and the source and date sit beside the number. Refresh due after India's large IT services companies report results for the quarter ending 31 December 2026, expected around mid-January 2027 (the dates are not confirmed here). This field moves by the quarter, so treat this as a dated scorecard, not a permanent answer.

If you only remember one line from this guide, make it this: a number about AI and jobs is only as good as its method, so before you share it, ask what was counted, who was asked, over what period, and whether the source can show cause or only association.

Six questions readers ask (quick answers)

• Is AI taking jobs in India? No source can answer that for the whole economy. The two most quoted studies point to AI-linked hiring ahead of AI-linked layoffs (Nomura) and no evidence of large-scale job losses in IT (ICRIER), and both show entry-level roles as the weak spot.
• What is the 2.6 to 1 ratio? Nomura's tally for India from 2022 to August 2026: 83,100 AI-related hires against 31,921 AI-linked layoffs and attrition, taken from company announcements reported in the news. It is a count of reported cases, not a census.
• Are freshers hit hardest? The entry-level signal shows up in three sources: Nomura, the ICRIER survey (55% of firms with falling entry-level employment, 35% rising) and foundit's July 2026 job postings (fresher postings down 10% year on year). It is not universal: Infosys says it still plans about 20,000 graduate hires this financial year.
• Do the second-quarter results of TCS (Tata Consultancy Services), Infosys and Wipro prove anything? They show three large employers moving in different directions in one quarter: net hiring of 4,258 at TCS and 4,480 at Infosys, and a fall of 6,716 at Wipro. One quarter cannot show cause.
• Which numbers should I trust least? Totals with no named cases or counting rules, projections of job postings read as jobs, and figures credited to a source that does not contain them.
• What should I do about it? Treat the headlines as weather, not a forecast for your own role. Map which of your tasks AI can do, build and measure one real piece of AI-assisted work, and keep the evidence.

Three forwards in one week

This is an illustrative scenario, not a real company or person. A mid-career analyst in Pune receives three screenshots in one week. The first says AI adds 2.6 jobs for every one it removes in India. The second says more than half of IT firms report falling entry-level employment. The third says two-thirds of AI workers expect layoffs in their team. Each is quoted correctly. Together they look like a contradiction, and the analyst cannot tell whether to learn, wait or worry.

They are not a contradiction. The first counts company announcements, the second counts firms that answered a survey, and the third counts how safe people feel. Each is a different kind of evidence about a different slice of the market. The scorecard below puts them side by side so you can see which part of each headline holds up.

The scorecard: thirteen numbers, graded

The main finding of the check is that very few numbers failed outright. Most were accurate and over-read. The grades work like this:

• Holds up. The number is in the original source, matches it, and measures what it is usually taken to measure.
• Holds up with limits. The number is real and traceable, but it measures something narrower or different from the headline reading: a survey of opinions, a count of announcements, a projection, or a derived figure.
• Cannot be traced. The original could not be found, or the number is not in the source it is credited to.

What the Nomura numbers can and cannot tell you

Nomura's economists, Sonal Varma and Si Ying Toh, started from a gap: labour market data in Asia are fragmented. So they aggregated public company announcements of job losses and new hiring that the companies tied to AI. Their sample is 69 announcements across Asia, mostly from 2022 to August 2026, and the chart note in their published piece splits it into 35 anecdotes on layoffs plus attrition and 34 on hiring. The written note gives the Asian totals: about 131,000 AI-related hires against about 61,000 AI-related layoffs. Bloomberg's coverage, republished by ThePrint on 7 August 2026, gives the India split: 83,100 hires against 31,921 layoffs and attrition.

The arithmetic is simple. India's 83,100 hires less its 31,921 layoffs and attrition is a net of 51,179, a ratio of about 2.6 to 1. Using Business Today's Asia totals of 130,758 hires and 60,654 job losses, India accounts for about 64% of the hires and about 53% of the layoffs and attrition. Dividing Nomura's Asian totals by its anecdote counts gives roughly 3,800 jobs per hiring anecdote and 1,700 per layoff anecdote. Averages hide the spread, but they show how few events sit behind each bar: one large announcement can move a country's total.

What it can tell you: among the cases that companies chose to attribute to AI, hiring outweighed cuts in India so far. Nomura's own summary reads: "So far, the net impact of AI has been positive for India." It then adds the sentence that matters most to a worker: displaced workers rarely move directly into AI engineering roles, so the aggregate may hide real distributional pain. In India, Nomura says, hiring has centred mainly on data annotation and linguistics, most firing cases were support teams replaced by chatbots, and most hiring cases were IT services graduates whom firms explicitly link to AI demand.

What it cannot tell you: Nomura says much of this itself. The sample is announcements, not a representative labour survey. "AI-related" means what companies publicly say it means, and the note does not define "attrition", which appears only in the chart label "layoffs + attrition". Silent cuts, such as hiring freezes, reduced campus recruitment and early retirement, are less visible than mass layoffs and harder to measure. Effects outside technology have not shown up yet. And a net of 51,179 does not mean 51,179 people are better off, because the hired and the displaced are different people with different skills.

What the ICRIER survey can and cannot tell you

The ICRIER report, "AI and Jobs: This Time Is No Different" (February 2026), by Deepak Mishra, Mansi Kedia, Aarti Reddy, Cledwyn Fernandez and Agrima Khanduri, is, by its authors' account, the first survey to collect evidence systematically on AI's effect on hiring, productivity and skilling in India's IT sector. It draws on 651 respondents in five segments (IT services, IT products, business process management, engineering services and research and development, often called Global Capability Centres or GCCs, and startups) across 10 cities, surveyed between November 2025 and January 2026, 72% of them in person. The authors say the survey is not fully representative of the sector, that it records the direction of change rather than precise magnitudes, and that in some large firms more than one respondent was interviewed, so 651 respondents is not necessarily 651 different firms. The acknowledgements thank OpenAI for supporting the study, and ICRIER's disclaimer says corporate funding does not imply endorsement of a sponsor's views or products.

The results are specific. On overall hiring, 65% of respondents said AI had decreased it, 24% said increased and 11% said no major change. On employment by level over the last 24 months, entry level had 55% of firms reporting a decrease, 10% no change and 35% an increase. Mid level had 25%, 33% and 42%. Senior level had 14%, 82% and 4%. Looking two years ahead, 27% expect their workforce to shrink, 28% expect it to grow and 44% expect no major change.

Three details change how to read those figures. First, they are shares of respondents, not numbers of jobs: a firm that dropped two trainees and a firm that dropped two hundred count the same. Second, the entry-level question as printed asks about the change in employment over 24 months, not about AI. Nomura's chart and many headlines paraphrase it as hiring, and the report reads it alongside AI adoption, but the question does not itself ask why. Third, the report's evidence that total IT employment kept growing comes from household survey data that ends in 2023-24, so it mostly predates the 24 months the survey asks about.

The authors are open about the limits of causation. A footnote offers an alternative explanation for the link between AI exposure and rising demand: firms may be applying AI to the tasks where demand is already growing fastest. The report also notes that the hiring slowdown matches a broader deceleration in revenue growth and employment expansion after the post-COVID boom. Their conclusion is that they found no evidence to support predictions of large-scale job losses after generative AI adoption in India's IT sector, and instead saw a modest, broad-based moderation in hiring concentrated mainly at the entry level. They read that as augmentation more than automation, and they say the conclusions may need revisiting if AI use expands dramatically.

What three results releases can and cannot say

TCS, Infosys and Wipro each reported the second quarter of financial year 2026-27, covering July to September 2026. Here is what each said in its own documents.

Add the three headcount changes and the net is plus 2,022 on a combined base of 1,166,926, or 0.17%. That is a fact about one quarter at three firms. It is not an AI effect and not a trend: the firms moved in different directions, and Wipro's release gives no reason for its 6,716 fall. Infosys's headcount is only 551 higher than a year earlier, which is a reminder that a quarterly gain can sit inside a flat year.

The AI revenue figures are real company statements, and they are not comparable. TCS gives an annualised run-rate in dollars, Infosys gives a share of quarterly revenue, and Wipro gives nothing. The widely repeated Infosys figure of US$540 million is not in Infosys's documents. It is 10.5% of the US$5,145 million quarterly revenue, which is US$540.2 million, and some secondary write-ups derive it that way. It is a quarterly number, so it cannot be set against TCS's annualised US$3.1 billion without adjusting for that. The documents we read do not define what counts as AI revenue at any of the three.

The guidance cut is a revenue number, not a headcount number. Infosys cut its FY27 growth guidance to 1% to 2% in constant currency from 1.5% to 3%. On its calls, management pointed to softer volumes, pricing pressure and competitive intensity, higher client demands for AI-led productivity, weakness in retail and hi-tech, and a difficult macro environment. Its chief executive, Salil Parekh, described AI as working in both directions: it is giving the company growth, and it is also seeing compression in the price of some work. In the same calls Infosys said it hired about 13,000 graduates in the first half and still targets about 20,000 for the year, the same as last year, and its chief financial officer said a fresher typically takes six to eight months to train and deploy. All of that sits in one quarter: AI lifts some revenue, AI lowers the price of some work, and the company is still hiring freshers.

Two softer kinds of number: job postings and feelings

foundit: a projection of postings in a falling market

The foundit Insights Tracker, from the online jobs platform foundit, is a monthly analysis of online job postings. Its December 2025 edition, published in January 2026, counted 290,256 AI job postings in 2025 and projected about 381,846 in 2026, a rise of about 32%. That is the source of the widely quoted line that AI hiring will grow 32% to about 3.8 lakh roles. Three limits matter. It counts postings, not jobs filled. It is a projection made in January, and foundit restated it in July 2026 as close to 3.8 lakh roles without a revised figure that we could find. And the same July 2026 edition reports that white-collar hiring activity fell 12% year on year, with 18 of the 26 industries it tracks lower and fresher hiring down 10%. The market around the AI roles is shrinking, and the report does not attribute that to AI: it describes employers moving from broad expansion to targeted recruitment.

The July 2026 edition also adds composition. Internships, contracts and project work made up 32% of entry-level openings in the first seven months of 2026, against 21% in 2025 and 12% in 2024. AI-led roles rose from about 1% of fresher postings to 4% in a year, and GCCs raised their share of fresher hiring from 8% to 11%, while IT services took a smaller share than before. These are shares of postings on one platform's index, but they match the pattern in the other sources: fewer conventional entry-level jobs and a small but growing set of AI-led ones.

Blind: how safe people feel

Blind Workplace Insights (Blind is an anonymous community app for professionals) published a survey on 6 August 2026. It asked 1,552 India-based professionals, in July 2026, how likely a layoff or significant headcount cut was in their own team over the next 3 to 6 months, and counted "very likely" and "somewhat likely" as at risk. Among AI and ML respondents, 66% were at risk. The same survey put sales and marketing higher at 68%, product and design at 65%, engineering at 58% and data and analytics at 51%. Leading with the 66% leaves out that sales and marketing scored higher. The page does not say how respondents were recruited, how many were in the AI and ML group, or what the margin of error is, and it measures perceived risk, not layoffs. It is a reading of sentiment among users of one app, and it is useful as that.

What none of these sources can show

• The whole-economy effect. Nomura counts company announcements, ICRIER surveys IT firms, foundit counts online postings and Blind asks users of one app. None counts every job in India, and the national survey data in the ICRIER report ends in 2023-24.
• Cause. Every source above is consistent with AI doing some of the work and other forces doing the rest. ICRIER points to the post-COVID slowdown, and foundit finds hiring lower in 18 of 26 industries, which is not an AI-specific pattern.
• Who is affected. A positive net can hide the fact that displaced support staff and newly hired data workers are different people. Nomura says as much.
• Job quality. A count of jobs says nothing about whether they are permanent or fairly paid. foundit's 32% of entry-level openings as internships, contracts or project work is a hint that the type of job is changing.
• Timing. Nomura says the full impact may take years to appear and that effects outside technology have not been seen yet. ICRIER says its conclusions may need revisiting if AI use expands sharply.

Where the sources agree: entry-level roles are where the pressure shows, and experienced or specialist roles are firmer (Nomura on senior talent, ICRIER on mid-level hiring, foundit on specialist skills). Where they differ: ICRIER reports no sign of large-scale IT job losses, Nomura reports net AI hiring but warns of distributional pain, and Blind shows people feeling at risk regardless. These are different kinds of evidence, and they do not have to match.

When the headline goes wrong: symptom, cause, fix

How to read the next headline: a ten-point checklist

1. Can you open the original? A headline is not a source. Find the report, filing or survey page.
2. What was counted? Hires, job postings, company announcements, opinions or payroll are five different things.
3. Is it a share or a count? "55% of firms" and "55,000 jobs" are not interchangeable.
4. Census or sample? If it is a sample, how was it chosen and how big is it?
5. Over what period? And is that period still current, or has a newer quarter changed it?
6. Does the question or method mention AI? Or is AI the writer's explanation for a change?
7. Observed or projected? A forecast made in January is not a measurement taken in October.
8. Who paid for it, or sells something connected to it? Know it, do not dismiss it. ICRIER's disclaimer on corporate funding is an example of how a source can state this itself.
9. Does a net figure combine different people? Hires minus cuts is arithmetic, not a trade-off any one person experienced.
10. What does the source say it cannot show? Read the caveat paragraph before the headline number.

Brief your AI assistant to grade a headline

You can use an AI assistant for the first pass, as long as you give it the text and forbid it from filling gaps. You supply the headline and the pasted source. The assistant extracts the fields, and you check the answers against the original, because an assistant will sometimes supply a plausible guess where the source is silent.

You are helping me check a headline number about AI and jobs. I will paste the headline and, if I have it, the text of the original report.
Use only the text I paste. Do not browse and do not add numbers from memory.
For the number, fill in these fields. If the pasted text does not say, write NOT STATED.
1. The exact number and the exact wording used in the source
2. Who produced it and on what date
3. What was counted (hires, job postings, company announcements, survey answers or payroll)
4. The unit: a share of firms or people, or a count of jobs
5. The period covered, and whether it is observed or projected
6. The sample: how many, how chosen, and whether the source calls it representative
7. Whether the source says AI caused the change, quoting the sentence
8. What the source says it cannot show
Then give a grade: holds up, holds up with limits, or cannot be traced, with one sentence on why.
Finish with the three questions I should ask the author of the headline.

Read the answers to fields 1, 4 and 7 against the source yourself. The NOT STATED rule reduces invented answers and does not remove them.

What to do about it

None of this tells you what will happen to your own job. It does show where the evidence points, and what is worth doing while the numbers stay unsettled. The steps below are our advice, not findings from these sources, and none of them guarantees a job, a promotion or a salary.

1. Treat the headline as weather, not a forecast for your role

Every source above describes averages across firms. Your exposure depends on your tasks. The Kill It or Keep It test is a quick way to sort the tasks in your own job into the ones AI can do, the ones it can help with and the ones that need you. For the wider picture of where AI is heading for Indian professionals, see AI Trends 2026.

2. Move toward the hybrid profile employers say they want

In the ICRIER survey, 63% of firms reported a shift toward candidates who combine domain expertise with AI and data skills, and 68% named prompt engineering and generative AI skills as in greater demand because of AI. Only 4% of IT firms had trained more than half of their staff in AI skills, so formal training still reaches a minority of staff at most firms, and building skills yourself still matters. How to Upskill in AI Without a Coding Background and the complete guide to learning AI cover where to start, and why a list of fifty tools is not a skill explains what to avoid.

3. Keep evidence of shipped work

foundit's July 2026 tracker describes employers hiring freshers on different terms and against a higher bar on demonstrable capability. Whatever your level, the response is the same: build one real piece of AI-assisted work in your job or a side project, measure it with a before and after number such as time saved, error rate or cost, and write down what failed. A short record of what you built, how you checked it and what you learned is stronger evidence than a certificate.

4. Prefer skills that travel across roles

In foundit's July 2026 data, the largest groups among AI-led fresher postings were AI and ML engineer (28%), data analyst (24%), prompt engineer (12%) and AI operations (10%), and the report describes employers as buying people who can put working systems into production and read what comes out. Nomura cautions that the long-term durability of entry-level data roles is uncertain, because evolving models may automate them too. Skills such as checking AI output against a test set, designing a workflow around a task and applying domain judgment travel across job titles better than any one title.

5. If you lead a team, plan the entry-level path on purpose

The pressure on entry-level hiring shows up in Nomura, ICRIER and foundit, and Infosys's chief financial officer noted that a fresher takes six to eight months to train and deploy, so a hiring decision made today shapes capacity months later. If you lead people, decide deliberately how new joiners will learn on the job when AI does the first draft. The AI adoption roadmap, the guide to change management for AI adoption and the AI roadmap for business leaders cover the planning side.

Frequently asked questions

Is AI taking jobs in India in 2026?

No source can answer that for the whole economy. Nomura's tally of company announcements shows AI-linked hiring ahead of AI-linked layoffs and attrition in India, 83,100 against 31,921. ICRIER's survey of 651 IT respondents found no evidence of large-scale job losses but reported entry-level employment falling at 55% of firms. Both are partial views, and neither counts every job.

How many AI jobs has India created?

No source gives a count of all AI jobs. Nomura's 83,100 is AI-related hires among company announcements, 2022 to August 2026. foundit counted 290,256 AI job postings in 2025, which are advertisements, not hires. They measure different things and cannot be added together.

What did the Nomura report say about India?

Nomura's economists said that, so far, the net impact of AI has been positive for India, with hiring ahead of cuts in the announcements they collected. They also described a two-tiered labour market, with falling demand for entry-level workers and rising demand for senior talent, and warned that displaced workers rarely move into AI engineering roles.

How reliable is the 2.6 to 1 ratio?

It is accurate as arithmetic on Nomura's data: 83,100 divided by 31,921 is about 2.6. It is weak as a measure of the labour market, because it rests on company announcements (69 across Asia), the terms "AI-related" and "attrition" are not defined, and silent cuts such as hiring freezes are hard to capture.

What did the ICRIER survey find about entry-level jobs?

Of 651 respondents in India's IT sector, 55% said entry-level employment in their organisation had fallen over the last 24 months, 35% said it had risen and 10% said it was unchanged. These are shares of respondents, not job counts, and the question did not mention AI. The authors describe a modest moderation in hiring concentrated at the entry level.

Is the ICRIER survey independent?

The report was written by ICRIER researchers and its acknowledgements thank OpenAI for supporting the study. ICRIER's disclaimer says corporate funding does not imply endorsement of the sponsor's views or products. The survey is not fully representative, by the authors' own account, so read it as a detailed sample, not a census.

Did TCS, Infosys or Wipro cut jobs because of AI?

The documents we read do not say so. In the quarter to 30 September 2026, TCS added 4,258 employees and Infosys added 4,480, while Wipro's headcount fell by 6,716 and its release gave no reason. One quarter at three companies cannot show what AI did to employment.

Is foundit's 32% AI hiring projection reliable?

It is a real figure from foundit's tracker, a projection of online AI job postings from 290,256 in 2025 to about 381,846 in 2026. It counts postings, not jobs filled, it was made in January 2026, and the same tracker's July 2026 edition shows overall white-collar hiring activity down 12% year on year.

How can I check an AI jobs headline myself?

Find the original report, then ask what was counted, who was asked, over what period, whether it is a share or a count, whether the source claims AI caused the change, and what the source says it cannot show. The ten-point checklist in this guide covers each step.

Which skills do Indian IT firms say are in more demand because of AI?

In the ICRIER survey, 68% of firms named prompt engineering and generative AI skills, 36% data analytics and 35% data science and machine learning. These are shares of firms in a multiple-choice question, and 63% of firms reported a shift toward candidates who combine domain expertise with AI and data skills.

How often will this guide be updated?

After the next quarterly results of the large IT services companies, which we expect around mid-January 2027, and sooner if a source revises its numbers. Each row of the scorecard shows the date of its source.

Build the evidence with us

To follow this series as each post publishes, join the Garage Labs community. To see where you stand today, take the AI readiness quiz. If you want to build an AI product copilot, a RAG (Retrieval-Augmented Generation) knowledge assistant with citations, an autonomous research agent and an evaluation harness, finishing with a capstone, those are the builds in the Applied AI Practitioner Program, taught through live sessions and build labs. The programme expects 5+ years of experience, and no course, ours included, can promise a job outcome. See the programme page for the current schedule and details.

Found this useful? Share it with someone who needs it. Visit garagelabstech.com and follow Sumit Kumar Singh for more on LinkedIn.

Sources and further reading

• Nomura Connects, Sonal Varma and Si Ying Toh, Beyond the Headlines: AI Has Added Jobs in Asia, So Far (August 2026)
• Bloomberg News, republished by ThePrint, AI-related hiring is outpacing job losses in India, so far, Nomura report (7 August 2026)
• Business Today, coverage of the Nomura report (9 August 2026), for the Asia totals
• ICRIER (Indian Council for Research on International Economic Relations), AI and Jobs: This Time Is No Different (February 2026)
• TCS, International Growth and Strategic Wins Underpin TCS' Q2 (8 October 2026) and first-quarter results release for the 30 June figures
• Infosys, second-quarter FY27 documents filed with the US Securities and Exchange Commission (SEC) on Form 6-K: press release, fact sheet, media call transcript and earnings call transcript (23 October 2026)
• Wipro, results release for the quarter ended 30 September 2026 and highlights data sheet, filed with the SEC on Form 6-K on 20 October 2026: release and data sheet
• foundit, Insights Tracker, December 2025 trends (January 2026) and Insights Tracker, July 2026 (August 2026)
• Blind Workplace Insights, 66% of India's AI workers brace for headcount cuts, while engineers feel safer (6 August 2026)