India's AI Job Disruption Map: 25 Roles, When They Change, and How to Build a Career That Survives

In May 2024, Andrej Karpathy — co-founder of OpenAI, former director of AI at Tesla, and creator of nanoGPT — published a small tool at karpathy.ai/jobs. He called it "a development tool for exploring BLS data visually." It is a treemap of 342 US occupations covering 143 million American jobs. Each rectangle is sized by total employment. You can colour the map by projected growth, median pay, education requirement — or, the most interesting one, Digital AI Exposure on a 0-10 scale.

The framework is brutally honest. A job that can be done entirely from a home office on a computer scores 7 or higher. Data entry and telemarketers hit 10. Roofers and landscapers sit at 0-1. Software developers and designers land at 8-9. Karpathy is careful: the score is "restructuring potential," not "elimination probability." It does not model demand elasticity, regulatory drag, or social preference.

The chart is fascinating because it is uncomfortable. You can see, at a glance, which slices of the US labour market sit closest to the AI fire. But it does not work for India. The US Bureau of Labor Statistics taxonomy maps poorly onto Indian work: India's economy is dominated by IT services, BPM, design, content, and outsourcing — categories that sit at the very top of Karpathy's exposure scale. India is, in a real sense, the country with the highest AI exposure per capita of any major economy in the world.

I have been writing about this shift for the past year — first about how businesses can become AI-ready, then about how individuals can start using AI productively, and most recently about how Google AI Mode is rewriting the rules of search. This article is the third piece of that picture: what the next five years look like for the Indian workforce specifically, role by role.

This article is the Indian equivalent of that map. I have built a smaller, focused version — 25 roles that together cover most of urban India's white-collar workforce. Each role has an AI exposure score on the same 0-10 scale Karpathy uses, plus a specific timeline (which year the disruption is most likely to peak) drawn from NASSCOM, McKinsey, Deloitte, Salesforce, and NITI Aayog data.

Then the harder part: what to do about it, with specifics. Where the new jobs are (and there are millions of them). Which adaptation strategies actually work. And the roles I personally believe will benefit, not suffer, from this shift.

Why India is the most AI-exposed labour market in the world

The US tech labour force is roughly 9 million workers across software, design, support, and content. India's IT-BPM workforce alone is 5.8 million per NASSCOM's FY2025 estimate. Add the broader services and outsourcing economy and the number climbs past 12-15 million. Almost all of this work is digital, repeatable, and English-medium — which is precisely the profile where modern LLMs perform best.

The composition is more concentrated still. According to NASSCOM's industry breakdown for FY2025:

• IT Services & ER&D: $55.7 billion in revenue — heavy on software, architecture, devops, QA
• BPM (Business Process Management): $54.6 billion — voice support, back-office processing, KPO, finance and accounting outsourcing
• Software products: $16.1 billion
• Hardware: $19.2 billion

Exports of IT services alone account for more than 65% of India's total IT exports. The customers are largely US and UK enterprises buying capacity at lower hourly rates than they could find at home. That entire pricing arbitrage is what AI threatens. If a US enterprise can deploy Claude or GPT-5 to generate equivalent output at $0.01 per task, the value of paying ₹500/hour for an Indian developer to write the same CRUD application drops sharply.

This is not a hypothetical. Outsource Accelerator's 2025 analysis projects 40% BPO job loss in India by 2028 as enterprises deploy AI-first support, and as much as 80% of Tier-1 support eliminated by 2028. Outlook Business reported in late 2025 that TCS cut 12,200 jobs (about 2% of its workforce) in July 2025, with the company directly citing AI-driven automation. Fresh engineer intake at major Indian exporters fell to 70,000-80,000 in 2024-25 — the lowest fresher intake in over two decades. The broader McKinsey Global Institute AI economic impact research models adoption scenarios where countries with large services-export workforces (India, the Philippines, Ireland) face the steepest near-term labour-market readjustment.

For founders and team leads watching this unfold inside their own businesses, the practical question is not "is AI coming" — it is "what do I automate first, and which roles do I redesign before the market forces my hand." I have written a detailed founder-level guide on what AI readiness actually means in 2026 if you want the operations-side version of this story.

That is the macro picture. The micro picture, role by role, is below.

The Indian AI Job Exposure Map — 25 roles, scored and timed

Here is the visualisation. Bar length shows the AI exposure score on the same 0-10 scale Karpathy used. Colour shows risk band: very high (9-10), high (7-8), medium (5-6), lower (under 5). The year next to each bar is when I expect disruption to peak for that role in the Indian market.

A few things to call out before going role by role. First, this is restructuring exposure, not extinction. A score of 9 means the work content is highly automatable, not that 90% of the people doing it are out of a job tomorrow. Second, the timeline column is when I expect peak Indian disruption — when pricing and hiring pressure are most acute. Third, India-specific factors (regulatory, language, customer trust, English-medium delivery to global clients) all delay or accelerate these timelines compared to the US.

Very high risk roles (score 9-10) — peak disruption 2026 to 2027

These are the seven roles where AI is already capable of doing most of the work today, and the only thing keeping headcount steady is enterprise inertia. Inertia clears fast once one big customer flips.

1. Data Entry / Document Processing — Score 10, peak 2026

The most automated role globally and the most automated in India. NASSCOM's own 2023 research found 67% of Indian BPO companies have already automated more than half of their data processing work. Modern document AI (OCR + LLM extraction) handles invoices, contracts, KYC forms, and medical records at near-human accuracy and under 5% of the per-document cost of manual entry. The remaining work is exception handling and quality assurance, which needs about 10-15% of current headcount.

2. Telemarketing / Outbound Calling — Score 9, peak 2026

AI voice agents (Bland AI, Vapi, ElevenLabs Conversational, Sarvam in Indian languages) are already running outbound campaigns in Hindi, Tamil, Telugu, and English at production scale for Indian fintech and edtech companies. The voice quality has crossed the uncanny valley for a 90-second sales conversation. Indian telemarketing employs roughly 600,000-800,000 people across BPO and direct-employer setups. Most of this headcount will be displaced or repurposed by end-2026.

3. Tier-1 Voice Support (BPO) — Score 9, peak 2027

India's voice support sector is roughly 1.4-1.6 million workers. Voice agents from Ada, Cognigy, Yellow.ai, and Verint already handle Tier-1 queries (balance check, status updates, basic troubleshooting) at quality matching or beating human agents on consistency. Industry analysts project 80% of Tier-1 support eliminated by 2028. The remaining roles shift to escalation handling, AI quality review, and AI training data labelling.

4. Email and Chat Customer Support — Score 9, peak 2027

Easier to automate than voice because text generation is more reliable. Salesforce State of Service 2025 projects 50% of all service cases will be AI-resolved by 2027, up from 30% in 2025. Major Indian e-commerce companies already see 68% of Tier-1 chat queries handled by AI without human escalation. Indian banks report 60-70% of initial customer queries handled by chatbots. Inside India specifically, WhatsApp has become the default customer-support channel for retail and services — and most businesses are now wiring AI directly into it. I covered the build-vs-buy pricing and risks in detail in my deep-dive on WhatsApp automation in 2026.

5. Basic Content Writers / Blog Writers — Score 9, peak 2027

The first wave of disruption hit in 2023-2024 when ChatGPT killed the "₹0.50 per word" content market overnight. The second wave is hitting now as enterprises build AI-first content pipelines that produce 100 articles a day at zero marginal cost. Indian content agencies that built on the cost-arbitrage model are downsizing or pivoting to editorial quality control. Writers with strong original thinking, brand voice, and subject expertise are still in demand at higher rates — the floor collapsed, the ceiling held.

6. General Translators — Score 9, peak 2027

Google Translate, DeepL, Claude, and GPT-5 produce general-purpose translation at quality matching or exceeding average human freelance translators for most language pairs, including all major Indian languages. The exception remains certified legal, medical, and literary translation — where accuracy, liability, and creative interpretation matter. Indian general-purpose translation freelancing is collapsing in price; specialist translation is holding.

7. Bookkeepers and Junior Accountants — Score 9, peak 2028

Tools like Tally Prime, Zoho Books, and QuickBooks already include AI-driven transaction categorisation, reconciliation, and basic reporting. The Indian bookkeeping market employs an estimated 1.5-2 million people serving SMBs. By 2028, AI-augmented platforms will handle 70-80% of routine bookkeeping for businesses under ₹10 crore revenue. Senior CAs and audit roles are largely unaffected; the affected layer is the data-capture and routine compliance staff.

High risk roles (score 7-8) — peak disruption 2027 to 2028

These are the roles where AI is becoming powerful enough to handle 60-80% of the work but where human judgement, accountability, and quality control still matter. Headcount in this band shrinks 30-50% by 2028, but pay for the remaining roles rises because the work shifts to higher-value tasks.

8. Junior Software Developers (CRUD work) — Score 8, peak 2027

This is the canary that everyone in Indian IT is watching. Tools like Cursor, Claude Code, GitHub Copilot, and Replit Agent already produce production-quality CRUD code, simple API integrations, basic React components, and routine bug fixes faster than the typical 1-2 year experienced developer. TCS's 12,200-person July 2025 layoff and the collapse in fresher intake to 70,000-80,000 across major exporters in 2024-25 are the leading indicators. By 2027, the bottom rung of the Indian IT pyramid will be 40-60% smaller than 2024. GitHub's 2024 developer experience research found that developers using Copilot completed coding tasks 55% faster than control groups — multiply that by every full-stack team in the country and the headcount math becomes uncomfortable.

If you are reading this as someone hiring a developer in 2026 rather than working as one, the pricing and skill landscape has shifted significantly. My breakdown on WordPress developer pricing in India in 2026 walks through what each tier actually delivers in the AI-augmented era, and the companion piece on how to evaluate developers covers the right questions to ask.

9. Manual QA Testers — Score 8, peak 2027

Indian QA is a 400,000-600,000 person sector. AI testing tools (Testim, Mabl, Functionize, plus LLM-driven test generation) automate test case authoring, exploratory testing, and regression suites. Manual QA roles collapse to AI-augmented QA engineering roles, with roughly 50% of current headcount needed at the senior end.

10. Junior Graphic Designers (template work) — Score 8, peak 2028

Canva AI, Adobe Firefly, Figma AI, Midjourney, and dedicated tools like Looka and Brandmark already produce serviceable logos, social media graphics, brochures, and basic brand assets. The Indian graphic design market employs roughly 400,000-500,000 people. The template-and-clone work that constitutes most of the bottom 50% of that market faces severe pricing collapse by 2028.

11. Basic Video Editors — Score 8, peak 2028

Runway, Descript, Opus Clip, and CapCut AI handle most of the work for short-form social content, podcast repurposing, and basic talking-head edits. Indian YouTube and reels production agencies are already pricing in AI workflows. Long-form, narrative, and high-production-value video editing remains a specialist craft.

12. Insurance Claim Processors — Score 8, peak 2028

Document understanding plus rules-based reasoning plus image analysis (for damage claims) is exactly what modern AI is built for. Indian insurance back-office processing employs 200,000-300,000 people. Major insurers (HDFC ERGO, ICICI Lombard, Bajaj Allianz) are deploying AI claim assessment that targets 60-70% straight-through processing by 2028.

13. Paralegals and Legal Document Reviewers — Score 8, peak 2028

Tools like Harvey, Spellbook, and Indian-specific platforms like SpotDraft already handle contract review, due diligence, and case law research at the level of a 1-3 year associate. The Indian legal back-office employs roughly 100,000-150,000 people in document review and paralegal roles, plus another 50,000-80,000 in LPO (legal process outsourcing) serving US and UK clients. Senior lawyers are largely safe; the entry layer compresses sharply.

14. SEO Content Producers (mass scale) — Score 8, peak 2027

The model of producing 200 articles a month for a single B2B client at ₹1,000-2,000 per article is dead. Mass-produced SEO content gets discounted by both Google's helpful-content updates and the AI Overviews that summarise it without a click. Indian SEO content shops are either pivoting to strategy, original research, and expert content — or shutting down.

Medium-high risk roles (score 7) — peak disruption 2028

15. Junior Data Analysts — Score 7, peak 2028

Natural language data analysis (Hex, Mode, Snowflake Cortex, ChatGPT Code Interpreter, Julius) lets non-technical users get answers from data without writing SQL. Junior analyst roles that mostly produce dashboards and ad-hoc queries face significant displacement. Mid and senior analysts who build modelling frameworks, design metrics, and drive strategic decisions remain valuable.

16. Tax Preparers (Individual Returns) — Score 7, peak 2028

Indian individual tax filing is dominated by small CA shops and platforms like ClearTax, TaxBuddy, and Quicko. AI-driven filing for salaried taxpayers and small businesses will erode the manual prep market by 2028, though complex business and audit work remains specialist territory.

17. Junior UX/UI Designers — Score 7, peak 2028

Figma's AI features, Galileo AI, Uizard, and Visily generate competent UI mockups, design systems, and prototypes from text prompts. The junior designer role compresses; senior designers and product designers focused on user research, brand strategy, and complex interaction design hold steady or grow.

18. Social Media Coordinators — Score 7, peak 2028

Content calendars, post copy, hashtag research, scheduling, and basic analytics are all AI-doable today. The coordinator role merges into a smaller team of strategy-led, AI-fluent marketers.

19. Recruiters / HR Operations — Score 7, peak 2028

Sourcing, screening, scheduling, and initial outreach are largely AI-handled by platforms like Eightfold, Skillate, Phenom, and Paradox. The Indian recruitment services sector employs roughly 200,000-300,000 people; the routine sourcing layer compresses by 40-50% by 2028.

Medium risk roles (score 5-6) — peak disruption 2028 to 2029

20-23. Mid-level developers, designers, WordPress developers, marketing managers

This is the band where roles get reshaped rather than replaced. A mid-level full-stack developer in 2029 spends 30-50% of their day orchestrating AI agents, reviewing AI-generated code, and writing architectural prompts — not hand-typing functions. The total team headcount drops because each developer produces 2-3x more output, but the role itself survives. WordPress developers doing clone-and-customise work face the most direct pressure since AI can generate template implementations rapidly. Custom WordPress development, plugin engineering, performance optimisation, and security work remain expert domains.

24. Project Managers (transactional) — Score 5, peak 2029

Status reporting, dependency tracking, and standup summarisation are AI-doable. PMs who own roadmap strategy, stakeholder alignment, and tradeoff decisions are unaffected. The role bifurcates sharply.

25. Senior Software Architects — Score 4, augmented not replaced

The top of the engineering pyramid is the safest position in Indian IT. Senior architects who can orchestrate AI tooling, design systems, and make tradeoff decisions are the most in-demand profile in 2026 — and 2030.

The good news — and it is genuinely good

This is the part nearly every "AI is killing jobs" article gets wrong. The Indian opportunity is enormous, but it is unevenly distributed.

1 million+ AI-specific jobs by 2026

NASSCOM projects more than 1 million AI-related job openings in India by end of 2026. The Deloitte-NASSCOM AI Talent report says demand will grow from roughly 600,000-650,000 in 2022 to over 1.25 million by 2027. The supply is dramatically short: only 16% of current IT professionals are AI-skilled per MEITY data, and only 45% of Indian graduates are considered employable in the digital economy.

$15.7 trillion economic uplift by 2035

McKinsey estimates AI could add $15.7 trillion to India's economy by 2035. Cloud alone could contribute 8% of India's GDP and generate 14 million jobs by 2026. Most of these are not direct "AI engineer" jobs — they are jobs in AI-enabled businesses, supply chains, customer-facing roles, and new categories that do not exist today.

Salary inflation for AI-fluent professionals

India's AI job market grew 40%+ year-on-year per NASSCOM, with 15-20% annual salary growth expected to continue. The 2026 salary picture for AI roles in India:

• Data Scientists: ₹6-12 LPA entry, ₹15-28 LPA mid, ₹32-55 LPA senior, ₹50-80 LPA lead
• AI/ML Engineers: ₹8-15 LPA entry, ₹18-35 LPA mid, ₹40-70 LPA senior, ₹60 LPA-1 Cr+ for principal/founding engineers
• Prompt Engineers / Applied AI: ₹6 LPA entry, ₹15-25 LPA mid, ₹25-40 LPA senior
• AI Product Managers: ₹20-40 LPA mid, ₹40-80 LPA senior
• ML Platform / Infrastructure Engineers: ₹25-50 LPA mid, ₹50 LPA-1.2 Cr senior

Hyderabad, Bangalore, and Pune offer the highest packages due to dense AI startup and enterprise activity. Indian product engineering teams at US-headquartered companies pay 25-50% above the Indian market median because they are pricing against the global shortage. If you are weighing where to base yourself for the AI decade, I have a separate breakdown of the best Indian cities for tech and AI talent covering Bangalore, Hyderabad, Mumbai, Delhi NCR, and Pune across cost, talent density, and ecosystem maturity.

India's structural advantages

India has four advantages most countries lack: the largest English-speaking technical workforce in the world, a young median age (28 years), a strong STEM education base, and a 30-year head start on global services delivery. These advantages translate poorly to "competing with AI" but extremely well to "deploying AI for global customers." The countries that win the AI services decade will be those with deep customer-implementation expertise, not just model-training expertise. India is structurally positioned to be that country.

The World Economic Forum's Future of Jobs Report 2025 reinforces this view: 86% of surveyed employers globally expect AI to transform their business by 2030, and 70% expect to hire for new AI-related roles — with India and Southeast Asia identified as the regions with the highest absolute net job-creation potential through AI deployment, even as they face the highest displacement risk in services categories.

The adaptation playbook — what to actually do

If you are in a high-exposure role, here is the honest 12-24 month plan that I recommend to friends, clients, and people who write to me on LinkedIn. It is not a magic prescription. It is the highest-probability path I have seen work.

1. Pick one frontier AI tool and reach expert level in 90 days

Not "I have used ChatGPT a few times." Expert level. Pick one — ChatGPT, Claude, Gemini, or Cursor for engineers — and use it for two hours every working day for three months. Build something with it. Break it. Learn its limits. The career advantage of being top 5% with one tool dwarfs being mediocre with five. My Complete AI Starter Guide walks through how to pick the right tool for your work and where to start with prompting.

2. Move up the value chain in your current domain

If you are a developer, shift from writing code to designing systems, reviewing AI output, and orchestrating agents. If you are a designer, shift from production work to strategy, research, and brand. If you are in support, shift from handling tickets to designing AI agent workflows, training data, and quality evaluation. The role above yours is generally safer than your current role.

3. Build a public portfolio that shows AI-augmented output

Recruiters in 2026 are not impressed by candidates who say "I know AI." They are impressed by candidates who show: "Here is a project I shipped in 4 weeks that would have taken 3 months without AI. Here is the prompt strategy I used. Here is what I learned about its failure modes." Publish on GitHub, write a Medium or Substack post, or just put it on your LinkedIn profile.

4. Take a serious free course — and finish it

The Indian government's FutureSkills Prime initiative, Google's AI Essentials track, IBM SkillsBuild, and Microsoft's AI fundamentals are all free, government-recognised, and high-quality. None of them are difficult to enrol in. The completion rate among Indian learners is under 12%. Finishing one course puts you in the top decile.

5. If you are early-career, optimise for joining an AI-native team

If you are 0-3 years into your career, do whatever you can to work at a team where AI tooling is the default. The compounding learning over the first 3-5 years of an AI-native career is enormous and largely unrecoverable later.

6. If you are mid-career, double down on judgement and accountability

The roles that survive are the ones where someone has to take responsibility for outcomes. That is a human function and will remain one for a long time. Move into roles where you own a number, a customer, or a quality bar.

7. Build optionality outside your day job

Even in safer roles, build a side project, freelance practice, or consulting line. The 2026-2030 economy will reward those with multiple income paths and disadvantage those with single-employer dependency. If your only income is one paycheck from one company in one of the high-exposure categories, that is structural fragility.

Frequently asked questions

What is Andrej Karpathy's US Job Market Visualizer?
A research tool at karpathy.ai/jobs that visualises 342 US occupations covering 143 million jobs. Each rectangle is sized by employment and can be coloured by Digital AI Exposure (0-10), median pay, growth outlook, or education. Karpathy explicitly notes high exposure does not mean job elimination — it means restructuring potential.

How big is India's IT-BPM workforce in 2026?
Roughly 5.8 million people per NASSCOM. The industry is on track for $300 billion revenue in FY2026, with BPM at $54.6 billion, ER&D at $55.7 billion. McKinsey estimates 280 million Indian workers face some automation exposure by 2030.

Which Indian jobs will AI affect first?
Data entry, telemarketing, Tier-1 voice and chat customer support, basic content writing, general-purpose translation, and routine bookkeeping. Disruption peaks 2026-2027. 67% of Indian BPO companies have already automated more than half of data processing per NASSCOM.

Will programmers and software developers in India lose their jobs to AI?
Junior roles will shrink dramatically — fresher intake at major Indian software exporters fell to 70,000-80,000 in 2024-25, the lowest in two decades. TCS cut 12,200 jobs in July 2025 citing AI. Mid and senior developers shift toward AI-augmented work rather than displacement. Net effect: 30-40% fewer juniors but more mid-senior orchestration roles.

How many new AI jobs will India create by 2026?
NASSCOM projects 1 million+ AI-related job openings in India by 2026. Deloitte-NASSCOM says demand will grow from 600K-650K in 2022 to 1.25M+ by 2027. McKinsey estimates 950,000 new AI-related industrial jobs by 2030 and 14 million cloud-related jobs by 2026. Only 16% of current IT professionals are AI-skilled.

What is the safest career path in India for the next 5 years?
Three patterns: roles that combine deep domain expertise with AI orchestration, roles requiring physical presence or genuine human relationships, and roles building or governing AI itself. Avoid being mid-skill in fully digital, repetitive workflows with no proprietary judgement.

How quickly should I reskill if I am currently in a high-risk role?
12-24 months. NITI Aayog projects 54% of Indian employees will need significant reskilling. The free Indian government programs (FutureSkills Prime, plus Google/IBM/Microsoft tracks) are the highest-ROI starting point. Pick one frontier AI tool, master it, build proof-of-work that shows AI-augmented output, shift toward roles that orchestrate AI.

Will Indian outsourcing and BPO collapse entirely?
No, but the shape changes radically. The $54.6 billion BPM segment will remain. Work shifts from voice/email transaction processing to AI-augmented exception handling, complex judgement workflows, and AI training/labelling itself. Companies that move first to AI-first delivery models capture the next decade. 2 million BPO job decline by 2030 is realistic for routine roles, offset partially by net new AI training and orchestration demand.