2026-06-17

The State of AI Agents in India (2026)

India is moving fast on AI agents — but on two very different tracks: IT services and GCCs rebuilding around agentic AI, and a sharper debate on jobs, sovereignty, and language.

TL;DR: India is moving fast on AI agents — but on two very different tracks. On one track, IT services giants and GCCs are rebuilding themselves around agentic AI and BFSI/e-commerce/telecom are showing real production wins. On the other, there's a sharper-than-global debate about jobs (especially entry-level IT/BPO roles), whether India is building its own agent stack or just consuming others', and whether the "Indian-language, voice-first, agentic commerce" opportunity will actually materialize at scale.

Executive Summary

IT services are reinventing themselves around agents. TCS, Infosys, Wipro, HCLTech, and Tech Mahindra have all launched agentic AI platforms (WisdomNext, Topaz, AI Force 2.0, ADMS, etc.) and are repositioning from headcount-based to outcome-based pricing — while cutting jobs (top 5 IT firms cut ~7,000 roles in FY26).
GCCs are leaning in hard. 58% of India's Global Capability Centres are already investing in agentic AI, with another 29% planning to scale within a year — and most want to move beyond "cost arbitrage" to owning entire processes.
What's working: BFSI (HDFC, ICICI), e-commerce (Flipkart, Myntra), and telecom (Jio, Airtel) all have concrete production deployments with measurable impact — fraud detection, customer service, network optimization, AI-generated product design.
What's overhyped / contested: the jobs story is bifurcated — entry-level IT roles reportedly down 20-25%, even as agentic-AI-specific job postings grew 300%+. And there's a real debate about whether India is building sovereign AI capability (Sarvam, Krutrim, IndiaAI Mission) or remains a downstream consumer/application layer while talent and IP migrate abroad.
Where it's heading: India-specific bets are forming around regional-language voice agents (Sarvam, Krutrim) and "agentic commerce" on UPI/ONDC rails — with India possibly leapfrogging the West on agent-to-agent payments given its existing digital public infrastructure.

Background / Context

Unlike the global narrative — which is mostly about enterprise productivity and customer service — India's AI agent story is inseparable from its IT services industry, which employs millions and is the country's largest export sector. Every major IT firm has launched an agentic AI platform in the last 12-18 months:

TCS: WisdomNext (GenAI orchestration platform, 150+ specialized agents across financial services, accounting, supply chain)
Infosys: ~500 agents built via Topaz, across ~4,600 AI projects covering 90% of its top-200 clients; partnered with Cognition (Devin) for autonomous coding agents
Wipro: 200 production-ready GenAI agents with Google Cloud across healthcare, banking, insurance, retail, manufacturing; also a launch partner for Google Agentspace and Salesforce Agentforce
HCLTech: "HCLTech Insight" (agentic manufacturing, built on Google Cloud) and "AI Force 2.0" enterprise agent platform
Tech Mahindra: Agentic Development & Modernization Services (ADMS)

At the India AI Impact Summit (Feb 2026), Infosys, Wipro, Tech Mahindra, and Persistent Systems all unveiled NVIDIA-powered agentic AI platforms — a clear signal that "agentic AI" has become the new competitive battleground for Indian IT, alongside government-backed efforts like the IndiaAI Mission (₹10,000+ crore outlay, 40,000+ GPUs at subsidized rates).

Key Findings

What's actually working

GCCs are the quiet powerhouse. 58% of India's GCCs are already investing in agentic AI (another 29% plan to within a year); top use cases are customer service (65%), finance (53%), operations (49%), and IT/cybersecurity (45%). Fortune 500 GCCs in India now have 126,000+ AI-aligned roles, and 92% of GCCs want to move beyond cost arbitrage to own entire end-to-end processes — a meaningful shift in how global companies use their India operations.
BFSI is the standout sector, consistent with the global pattern of customer-service/ops wins translating well to India. HDFC Bank is targeting "AI-first enterprise" status within two years, using agentic AI for cybersecurity and routine task automation; its chatbot Eva handles millions of queries. ICICI Bank uses AI across loan processing, fraud detection, and voice banking via its iPal assistant. India's AI-in-BFSI market is projected to grow from $830M (2024) to $8.09B by 2033.
E-commerce and telecom have concrete, named deployments: Flipkart built "AgentServe" (gen-AI agents assisting customer service reps, reducing call handling time, lifting incremental GMV) and acquired a majority stake in Minivet AI for AI-led product discovery. Myntra's "Moda Rapido" is described as India's first fully AI-automated apparel design collection with zero human intervention in design creation. Reliance Jio uses agentic AI for real-time network optimization; Airtel's network-level AI has blocked 26B+ spam calls/SMS.
Indian leaders are bullish on near-term adoption: 93% of Indian business leaders plan to use AI agents within 12-18 months (Microsoft Work Trend Index 2025), and 24% are already deploying agentic AI in production (EY's "AIdea of India 2026" survey).

What's overhyped / contested

The "AI is taking jobs" story is real but more nuanced than headlines suggest. Entry-level IT roles have reportedly declined 20-25% as companies shift from mass fresher hiring to specialized hiring, and overall tech workforce growth slowed to just 2.3% in FY26. At the same time, agentic-AI-specific job postings (skills like LangChain, CrewAI) grew 300%+ in just over a year, and India is estimated to need 50,000+ specialized agentic AI professionals by 2027 — against a broader AI talent shortfall (India needs ~1.2M AI professionals by 2027 vs. ~420,000 in supply). The honest summary: routine roles are shrinking while a smaller, specialized pool of agentic-AI roles grows — a real inequality/transition problem, not a simple "robots took the jobs" story.
Integration and governance remain the top blockers, same as globally but more acute: 78% of Indian enterprises struggle with system integration, and 64.5% rate data governance/security as a "very severe" concern (EY).
Multilingual support is a genuine gap, not just a hype talking point. Inc42 reporting notes that global models still struggle with Hindi, Tamil, Telugu, Marathi, Bengali, and especially Hinglish/code-switching and incomplete sentences — meaning the "agents for the next billion users" vision isn't solved yet, even though it's the most-cited India-specific opportunity.
The "is India building or just consuming?" debate is unresolved. Bullish takes point to Sarvam AI (in talks to raise $250-350M at a $1.5B valuation, building India's sovereign LLM under the IndiaAI Mission) and Krutrim's "Kruti" (India's first agentic AI assistant, 13+ Indian languages). Skeptical takes argue India has no counterpart to OpenAI/Anthropic/DeepMind or to agentic coding platforms like Cursor/Cognition, and is a net exporter of AI talent — meaning India risks staying in the "application layer" while IP and profits accrue abroad.

Where things are heading

Agentic commerce on India's digital public infrastructure (UPI/ONDC) could be a genuine India-specific edge. Razorpay partnered with NPCI to demo AI agents that discover, compare, and complete purchases via UPI; Pine Labs launched a protocol (P3P) for AI agents to complete UPI payments without per-transaction human authentication; Cashfree and PayU launched similar "agentic payments" features. ONDC has crossed 200M cumulative transactions. A speaker at the India AI Impact Summit framed the opportunity as needing "an Aadhaar for AI agent identity, a UPI for inter-agent money flows, and an ONDC for interoperable commerce."
Voice-first, regional-language agents are scaling fast. Sarvam AI's "Listen at Scale" reached 5 million users in one month (March 2026), targeting farmers and healthcare workers in regional dialects. India's voice AI market is projected to grow from $153M (2024) to $957.6M by 2030.
IT services are betting their business model on this shift. TCS's annualized AI revenue crossed $2.3B in Q4 FY26; Microsoft announced 200,000+ Copilot licenses across TCS, Infosys, Wipro, and Cognizant. Infosys cites a concrete outcome example: 50 agents deployed for BP achieving 95% payment accuracy and 50% faster contract validation — illustrative of the "services to outcomes" pricing shift the industry is betting on.
Talent/skilling response is ramping up: government-backed initiatives include 15,000 scholarships for AI skilling in media/gaming/OTT (MIB + IICT + Google + YouTube), the "Srijan" GenAI Centre at IIT Jodhpur, and the "YuvAI" initiative with AICTE — all aimed at building the specialized agentic-AI workforce the market is now demanding.

Implications for PMs / Practitioners

If you're building for India, the BFSI/GCC playbook is the proven path: clear ROI metrics, integration with existing core systems, and a focus on fraud/compliance/customer service — the same categories driving adoption globally, just with an Indian enterprise's specific integration and governance constraints layered on top.
Don't underestimate the multilingual/voice opportunity, but don't oversell it either. It's the most India-specific differentiator (and Sarvam/Krutrim are proof it can scale), but code-switching and regional dialects remain real technical gaps — validate on real Indian-language data before committing.
Agentic commerce on UPI/ONDC is worth watching closely — it's one of the few areas where India's existing infrastructure (UPI's ubiquity, ONDC's open network) could let it move faster than markets without that digital public infrastructure.
If you work in or with Indian IT services, the "services to outcomes" shift is not optional — TCS, Infosys, Wipro, HCLTech, and Tech Mahindra are all repricing around agent-delivered outcomes, and clients are increasingly asking for it.
The jobs conversation needs nuance in any India-focused content. "AI is taking jobs" and "AI is creating a new specialized job category" are both true at once in India right now — useful framing for content aimed at a PM/tech audience that includes people on both sides of that shift.

Sources

Note on sourcing: several figures come from corporate press releases (TCS, Infosys, Wipro, HCLTech) and should be read as vendor framing rather than independent verification. The jobs-impact figures (entry-level role declines, NASSCOM growth numbers) come from different reports with different methodologies — treat the directional story (bifurcation between routine and specialized roles) as more reliable than any single percentage.