Indians Are Training the Robots That Will Replace Them: The ₹250-an-Hour Paradox

Disclosure: This post contains affiliate links to Amazon and other partners. If you buy through these links, digmod.com earns a small commission at no extra cost to you. We only recommend resources we genuinely think are useful.

Picture this. Nagireddy Sriramyachandra, 25, stands in her Chennai kitchen with a smartphone strapped to her forehead. She picks up a mango. Slices it. Drops the pieces in a steel bowl. The camera records everything from her point of view.

For one hour of this, she earns ₹250. About $3.

That footage gets sent to Objectways, an AI data company headquartered in Tamil Nadu’s Karur district. From there it flows to Fortune 500 clients, including teams connected to Amazon SageMaker. And eventually, it becomes training data for humanoid robots that will, one day soon, slice mangoes in someone else’s kitchen. Without needing a salary. Without taking breaks. Without needing ₹250 an hour, or anything at all.

Her colleague Kavin, 27, said it plainly to a recent AFP reporter: “In five or 10 years, they’ll be able to do all the jobs and there will be none left for us.”

This is the most uncomfortable trade happening in India right now, and almost no one is talking about it honestly. So let’s do that.

Inside Karur: India’s new “hand farms”

Walk into the Karur Objectways facility and you’ll see something strange. A textile factory floor where workers iron cloth bags and attach labels to caps, the kind of work Karur has done for decades. But interspersed among them, AFP photographers in May 2026 saw eight people wearing head-mounted cameras and smart glasses, doing micro-tasks: folding towels, picking up colored blocks, sorting utensils, slicing fruit, plugging cables. Each motion captured frame by frame from a first-person view.

The industry term for this is “egocentric data.” First-person footage, captured from the wearer’s point of view, used to teach AI systems how to move and act like humans in the real world. Chatbots and image generators crunch text and pictures. Embodied AI, the kind that lives in a humanoid robot, needs to see hands. Specifically, your hands.

One Objectways batch sent recently to a US client: 200 towel-folding clips, each one scrutinized for “silly errors” like an improper grip that might confuse the algorithm. Each video follows a strict choreography. Pick up the towel from a basket on the right, using only the right hand. Shake it straight using both hands. Fold neatly three times. Place it in the left corner of the desk. Miss a step or take longer than a minute, and you start over.

Bangalore-based Humyn Lab runs a similar operation. iMerit, another major Indian data annotation company, partnered with a stealth-mode US robotics startup to record, annotate, and classify 200 hours of in-home task data, using Meta Quest 3 head-mounted cameras and breaking the footage into 9 core household task types with 37 sub-classifications.

This is no longer a side experiment. It’s a supply chain.

Who exactly are these AI trainers?

The workforce splits across roughly three layers.

The video-capture workers. Mostly young, often women, frequently housewives or part-time workers from cities like Chennai, Bangalore, and across Tamil Nadu. They wear the cameras, perform the tasks, and submit the footage via apps. ₹250 an hour, no fixed schedule, paid per assignment. “Who else will give you 250 rupees an hour just for doing housework?” Sriramyachandra asked AFP from her Chennai kitchen.

The annotators. Slightly more skilled office workers who take the raw footage and tag every frame: which object is being held, which sub-task is being performed, where the hands are at each moment. This is where the dataset becomes machine-readable. Wages here are higher, but the work is repetitive and high-volume.

The rural microtask workforce. Companies like Karya, a social enterprise based in India, have built an entirely different model. Karya breaks AI data work into microtasks distributed via a mobile app to workers in rural areas, many of them low-income. They pay nearly 20 times the Indian minimum wage and have distributed over 40 million tasks to more than 40,000 workers across India so far. The ethical positioning matters here. Karya is the rare example of someone trying to do this without exploitation.

Add to this the engineering layer (the ML engineers in Bangalore and Hyderabad actually building the model pipelines) and you have a four-tier Indian AI workforce. Three of those four tiers are doing work that AI itself is on track to absorb.

The money math (and why it stings)

Here’s where the irony gets sharp. Look at the gap between what an Indian trainer earns and what the resulting AI capability is worth.

Let that sit for a second. The hourly rate paid to the human teaching the robot is a tiny fraction of the value that one finished robot will eventually create over its operational life. The trainers are essentially selling, very cheaply, the right to use their hand movements as intellectual property. The asset they own (their physical motion) becomes someone else’s permanent asset.

And once the model is good enough, that worker is no longer needed in any layer of the chain. Not as a video capturer. Not as an annotator. And not as the eventual housekeeper, factory hand, or warehouse picker the robot will become.

The white-collar carnage already underway

The cleanest place to see the future of AI displacement in India is not in Karur. It’s in the Bangalore offices of TCS, Infosys, and Wipro.

TCS cut 23,460 employees in FY26. The largest workforce reduction among Indian IT majors, ever. The company called it a pivot to an “AI-first services model” and a reduction of bench strength per client engagement. TCS share price dropped roughly 22.84% over the past year despite a strong Q4, reflecting market uncertainty about how the transition lands.

Infosys has run multiple consecutive rounds of layoffs in FY26, each one labeled “performance-based” rather than mass layoff. Around 700 freshers were let go last year for failing internal assessments three times in a row. The company filed a US WARN notice in February 2026 for 248 BPM employees. Wipro reduced fresher hiring guidance to 7,500 to 8,000, well below earlier estimates, and around 200 recruits publicly flagged that their onboarding had been deferred for more than 7 months.

The pattern is clear and worth repeating. Reuters reported expert estimates that TCS layoffs could be the start of a trend eliminating roughly half a million jobs from India’s $283 billion outsourcing sector over the next 2 to 3 years. WARN notice filings in the US by Indian IT firms accelerated sharply in Q1 2026 versus just 4 such filings in all of 2024. Bench periods at major firms have compressed from 30 to 60 days to around 15 days, and staying on bench beyond that increasingly leads to termination.

The labor arbitrage model that built India’s IT services industry, which roughly means “US firms pay Indian firms to do the same work cheaper with more people,” is breaking. If an AI agent can do the work of 10 BPO analysts, the cost advantage of having those analysts in Bangalore versus Philadelphia disappears. The premium is no longer in cheap labor. It’s in skilled, AI-fluent labor that can deliver outcomes with fewer humans.

Who’s actually at risk (tier-wise)

Most “future of work” coverage paints with too broad a brush. Let’s get specific.

Highest risk over the next 2 to 3 years

• BPO and call center agents (tier-1 support, especially in English)
• Data entry and document processing
• Content moderation (the major AI training pipelines are themselves automating this)
• Junior software engineers doing routine CRUD work
• Paralegals and contract review staff
• Junior accountants doing bookkeeping and basic reconciliation
• Data labelers themselves (yes, the irony, and it’s already happening)

Moderate risk over the next 3 to 7 years

• Mid-level software developers, especially in maintenance-heavy roles
• Radiologists and pathologists (image-heavy diagnostic work)
• HR coordinators
• Generic SEO content writers
• Translators for major language pairs
• Junior financial analysts

Relatively safer (for now)

• Skilled trades: electricians, plumbers, carpenters, mechanics
• Healthcare touch work: nurses, physiotherapists, ASHA workers
• Teachers, especially of young children
• Creative judgment roles: senior strategy, brand direction, design leadership
• Trust-heavy work: financial advisors, sales for complex B2B products
• Coordination roles that require deep context: engineering managers, account directors

One sober note. “Relatively safer” does not mean safe. The reason the trades feel safer right now is that humanoid robots are not yet good enough at general motion in unstructured environments. The workers in Karur are explicitly closing that gap. The honest answer about today’s skilled trades is that they’re safe for a decade, maybe more. Not forever.

What NITI Aayog says vs. what’s actually happening

The Indian government is aware. NITI Aayog released a study in October 2025 titled “AI for Inclusive Societal Development,” developed in partnership with Deloitte. Some of the key framing in that report:

• “Most discussions around artificial intelligence and labour focus on white-collar professionals and predict an almost certain loss of jobs in the segment” without urgent action.
• “Little attention, if any, is paid to how AI can serve India’s 490 million informal workers.”
• If nothing changes, informal workers’ average annual income could stagnate at around $6,000 by 2047. The per-capita figure India needs to hit high-income status is around $14,500.
• Proposed solution: a national mission called Digital ShramSetu. The idea is to formalize India’s informal workers digitally, give them verifiable identities, ensure timely payments via smart contracts, and connect them to social security benefits.

The framing is hopeful and the goals are right. The implementation timeline is years away. The TCS layoffs are happening now. The trainees on Karur factory floors are working now. If you’re a working Indian today, the realistic plan is to assume policy will lag your career by at least 5 years, and prepare accordingly.

Self-assessment: your personal AI risk score

Score yourself on each of the six questions below. Add the totals. Your score range tells you where you stand.

Reading your score:

• 0 to 5: Low risk. Your role mixes judgment, relationships, and novelty. Keep building AI literacy as a force multiplier.
• 6 to 10: Medium risk. You’re in the safe lane for now, but the lane is narrowing. Start AI fluency work this quarter.
• 11 to 14: High risk. Your role overlaps significantly with what AI is automating. Start a serious upskilling and side income plan in the next 30 days.
• 15 to 18: Critical risk. Treat the next 6 to 12 months as an active career transition window. Don’t wait for the layoff. Pre-empt it.

5 misconceptions to drop

1. “Only blue-collar jobs are at risk.”
False. TCS cut 23,460 white-collar IT jobs in FY26. Infosys, Wipro, HCL are all restructuring. Junior software roles, paralegals, junior accountants, content moderators are at higher risk right now than most skilled trades.

2. “AI will create more jobs than it destroys in India.”
Not at the rate India needs. India needs roughly 78.5 lakh new non-agricultural jobs every year. AI-augmented roles being created so far are nowhere near that scale, and they require specialized skills most displaced workers don’t have.

3. “Freshers are safer because they’re cheap.”
The opposite. Freshers do the work AI does best. Infosys laid off 700 freshers last year. Wipro deferred onboarding by 7 plus months for many. The cheapest ₹2 lakh per year employee now competes directly with a $20 per month AI subscription.

4. “The government will protect us.”
NITI Aayog has a real roadmap. Digital ShramSetu is being designed. But implementation is years away. The IT layoffs are happening now. Don’t bank on policy timing matching your career timing.

5. “I can wait this out.”
The compounding rate of AI capability is too fast. The you of 2026 is still in the labor market in 2030. The robots being trained today in Karur are projected to be in commercial deployment by 2028 to 2030. The window to act is now, not after the next round of layoffs.

What you should actually do (six moves)

If you read this far and feel a little anxious, good. Use it. Here’s the practical playbook.

1. Build AI literacy now, not later.
The single highest-return investment of your next 6 weeks. Anthropic Academy, Google AI Essentials, and Andrew Ng’s DeepLearning.AI short courses are free. Spend 4 to 6 weeks on this. Reduce your scrolling time by half for two months and you have the hours.

2. Become the prompt engineer at your current job.
Don’t wait for your company’s “AI strategy” to land. Start using Claude, ChatGPT, or Gemini in your daily workflow tomorrow. Document the tasks you automate and how. This positions you as the person who deploys AI, not the person AI deploys against. Internal AI champions are the last people any company lays off.

3. Move up the value chain in your current role.
If you do routine execution work, push for judgment work. If you write code, learn architecture and review. If you do tier-1 support, move to escalation handling. The most repeatable parts of your current job are the parts AI will absorb first. Get out of those parts before they get you out.

4. Pick one AI-resistant deep skill and layer it on.
Something that needs hands, trust, or deep context. Examples: financial advisory, complex B2B sales, teaching, healthcare allied roles, creative direction. You’re not abandoning your current career, you’re building optionality on top of it. Plan for 12 to 18 months of evening and weekend investment.

5. Build a side income today, while your salary is still steady.
This is the firewall. Even ₹15,000 to ₹25,000 per month from a side income makes you dramatically less desperate if your main job becomes shaky. Affiliate income, freelance work, digital products, consulting. The right time to build this is when you still have a paycheck and feel no pressure. If you’re new to this, our complete affiliate marketing guide for Indian beginners is the obvious place to start. Pair it with our email list growth post so you have an audience to sell to, and look at our first-time investing guide to put that extra income to work.

6. Get out of pure execution roles.
If your job description could be written as a clear set of repeatable steps, that’s the same job description AI is reading. Move toward roles that involve relationship, judgment, or coordination. The hardest jobs to automate are the ones where the boundaries of the job itself are fuzzy.

While you’re at it, audit how you spend the rest of your time. The single biggest threat to all of this is not AI. It’s procrastination. The plan is straightforward. The execution is hard. Use techniques like the Pomodoro method to protect daily upskilling time the way you’d protect a job.

Books, courses, and starter kits

If you want to go deeper, here’s a curated stack. Indian context first, then global picks.

Premium kits (built for Indian readers)

• AI-Ready Career Switch Kit by digmod (on InstantBundles): 60-day upskilling roadmap, AI literacy checklist, prompt template pack for office work, resume rewrite framework for AI-era roles. Built specifically for the Indian job market context. View on InstantBundles.

Books to read (Amazon)

• Co-Intelligence: Living and Working with AI by Ethan Mollick. The single most practical book on using AI at work right now. Read this first.
• AI 2041: Ten Visions for Our Future by Kai-Fu Lee and Chen Qiufan. Long-term scenario thinking on AI and work, written partly by a former Google China head.
• The Coming Wave by Mustafa Suleyman. What the next decade of AI plus biotech looks like, by a co-founder of DeepMind.
• Power and Prediction: The Disruptive Economics of Artificial Intelligence. The economist’s view of how AI changes business and jobs.

Free resources

• Anthropic Academy. Free AI courses from the makers of Claude.
• Google AI Essentials on Coursera. Free 10-hour intro for non-technical professionals.
• DeepLearning.AI short courses. Andrew Ng’s free deep-dive courses.
• NITI Aayog. The full “AI for Inclusive Societal Development” report is available as a free PDF on the NITI Aayog website.

Ready-to-use AI prompts for your career

Copy any of these into Claude, ChatGPT, or Gemini. Fill in your context. The output is genuinely useful career planning, free, in 5 minutes.

You are a senior career coach with deep knowledge of how AI is reshaping job roles in India. I will paste my LinkedIn "About" section and current job description below. Do three things:

1. Identify the parts of my role most exposed to automation in the next 3 years.
2. Suggest specific skill additions that would make my profile more AI-resilient.
3. Rewrite my "About" section to position me as an AI-fluent professional in my domain.

Be specific. Use Indian job market context. Don't use buzzwords.

[Paste your About section here]
[Paste your current job description here]

I work as a [role] in [industry] in [city] with [X] years of experience. My current monthly income is around ₹[amount]. I can invest 8 to 10 hours per week into building a new skill on top of my current work.

Give me a ranked list of 5 specific skills I could realistically learn over 12 months that:
- Are not easily automated by current or near-future AI
- Would let me layer this on top of my current career (not replace it)
- Are in real demand in the Indian market in 2026

For each, give me: why it's AI-resistant, expected income range in India, and one concrete first step.

I want to go from zero AI literacy to genuinely useful AI fluency in 60 days. I have 1 hour per weekday and 3 hours total on weekends.

Design my plan as a week-by-week schedule. For each week, give me:
- Specific topic to learn
- One free resource (course, article, or video) to use
- One concrete daily exercise I can do in 20 minutes
- One way I should practice the skill in my actual job that week

I work as a [role] and use [main software tools] daily.

FAQ

Are Indians really training the robots that will replace them?
Yes. Companies like Objectways in Tamil Nadu pay workers around ₹250 per hour to film themselves doing household and factory tasks. The footage trains humanoid robots being built by US tech firms, including clients connected to Amazon SageMaker. Multiple workers interviewed by AFP have said openly that they expect those robots to eventually replace their kind of work.

Which Indian jobs are safest from AI?
Skilled trades like electricians, plumbers, and carpenters, healthcare touch work like nurses and physiotherapists, teaching of young children, and senior judgment roles like financial advisory and complex B2B sales are relatively safer for now. No job is fully safe in the 10 plus year horizon.

Can I really upskill in 6 months?
Yes, for foundational AI literacy and prompt engineering. You will not become a senior ML engineer in 6 months. But you can absolutely become “the person at my company who uses AI well” in that time, which is enough to protect most mid-career professionals.

Is TCS still hiring freshers in 2026?
TCS has said it plans to hire around 40,000 freshers in FY27. But it also cut 23,460 jobs in FY26 and is pivoting to an AI-first services model. The hiring is real, but the bar is higher and the skills required have shifted heavily toward AI, cloud, and data work.

How many Indian IT jobs could be lost to AI?
Reuters reported expert estimates of around 500,000 jobs eliminated from India’s $283 billion outsourcing sector over the next two to three years. The World Bank has separately estimated that up to 69% of jobs in India are at risk of automation over the longer term, with risk concentrated in repetitive low-skilled work.

What is “egocentric data” and why is it valuable?
Egocentric data is first-person video footage captured from the wearer’s point of view using head-mounted cameras or smart glasses. AI developers use it to teach robots to copy human actions instead of programming each motion by hand. Companies like Objectways and Humyn Lab specialize in producing this kind of data in India.

How much do AI data trainers in India actually earn?
AFP reporting shows workers at Objectways earn around ₹250 (about $2.40 to $3) per hour for recording task videos. Karya, a social enterprise, pays its workers nearly 20 times the Indian minimum wage and has distributed over 40 million tasks to more than 40,000 workers, mostly in rural areas.

Will the Indian government step in to protect jobs?
NITI Aayog has released a roadmap called “AI for Inclusive Societal Development” and proposed a Digital ShramSetu mission to formalize India’s 490 million informal workers. Implementation timelines are long. Practical job protection from this policy in the next 2 to 3 years is unlikely. Plan as if the policy timeline does not match your personal career timeline.

Are humanoid robots really going to be deployed at scale?
Morgan Stanley projects more than 1 billion humanoid robots in use globally by 2050, mostly in industrial and commercial settings. Goldman Sachs sees the humanoid robot market hitting around $38 billion by 2035. Tesla, Figure AI, and others are already running pilot deployments today.

What should an Indian software engineer specifically do today?
Start using Claude, ChatGPT, or Gemini daily at work. Document what you automate and the time saved. Move into AI-aware roles inside your company such as AI integration, prompt engineering, or AI safety review. Pick up one creative or judgment-heavy skill on the side. Build a side income. Stop assuming your current role will exist in its current form in 5 years.