Agentic AI and the Future of Work in Service

Expert SpeakRaisina Debates

Published on Jul 21, 2026

As agentic AI compresses routine service work, the divide will run between economies that govern it and those merely exposed to it.

The global business process outsourcing (BPO) industry has hit a wall. For two decades, emerging economies rode a wave of cheap fiber-optic cables and wage arbitrage. They built massive industries on a simple premise: human labour in the Global South countries is cheaper than in the Global North countries. That premise is dying fast.

The culprit is Agentic AI. Unlike the chat assistants of 2023 that merely summarised text, today’s AI agents execute multi-step workflows across complex corporate systems. They authenticate users, query databases, file compliance forms, and issue refunds. They do it in seconds. And they do it for pennies.

To survive, service-exporting nations must pivot. The old model of third-party transaction processing is a race to the bottom. The future lies in Global Capability Centres (GCCs) – captive, highly integrated hubs that manage proprietary corporate intelligence rather than commoditised tasks.

The macroeconomic stakes are enormous. According to the IT and Business Process Association of the Philippines (IBPAP) in January 2025, the country’s outsourcing sector generated US$38 billion in 2024, employing 1.82 million people and contributing 8.5 percent of GDP, with revenues projected to reach US$59 billion by 2028. India’s numbers are even larger. NASSCOM’s Strategic Reviews show India’s technology sector generated US$282.6 billion in revenues in fiscal year 2025, a 5.1 percent increase, and crossed US$315 billion in fiscal year 2026.

To survive, service-exporting nations must pivot. The old model of third-party transaction processing is a race to the bottom. The future lies in Global Capability Centres (GCCs) – captive, highly integrated hubs that manage proprietary corporate intelligence rather than commoditised tasks.

Globally, GigaBPO’s 2026 market data values the outsourcing market at US$320 billion, with projections reaching US$696 billion by 2033. But these revenues are highly exposed. A 2024 IMF study estimated that AI exposure affects 40 percent of jobs globally, rising to 60 percent in advanced economies. For countries built on service exports, this is a macroeconomic emergency.

Task Compression, Value Migration, and the Human Face of Automation

Three forces shape AI’s economic impact on outsourcing: task compression, value migration, and workflow automation.

Task compression shrinks the hours required per job. When Klarna deployed its OpenAI-powered customer service assistant in early 2024, the bot handled two-thirds of customer chats in its first month, the equivalent of 700 full-time agents. Tier-1 BPO operators had to rapidly retrain thousands of agents who previously handled simple password resets into complex Tier-3 technical support and risk management roles.

As routine tasks compress, economic value migrates from volume-driven jobs to governance-intensive exception handling. At major logistics firms like DHL, automated document parsers replaced manual data entry. DHL partnered with HappyRobot to deploy AI agents to automate routine operations. When the parser failed on handwritten manifests, human operators stepped in as “exception handlers,” correcting OCR errors in real time to keep the supply chain moving.

The shift from generative AI (AI-assisted) to agentic AI (AI-enabled) accelerates this transition. While generative models draft emails, agentic AI runs the whole workflow. In a contact centre, an AI agent can now authenticate a caller, retrieve records, apply company policy, issue a refund, and update the CRM without human intervention. The shift from advisory to action increases efficiency but demands strict human oversight.

Divergent Realities: Advanced Economies vs. Developing Hubs

The automation shock hits different markets differently. In the UK, the impact has been acutely disruptive. A February 2026 Morgan Stanley AlphaWise survey of corporate executives revealed that UK firms experienced an 8 percent net job loss linked to AI over the preceding 12 months, the highest rate among peer nations. The UK Department for Science, Innovation and Technology (DSIT) noted in early 2026 that the country’s highly service-oriented economy leaves 70 percent of its workforce exposed to AI. Conversely, US companies reported a 2 percent net gain in jobs over the same period, with AI-related hiring outpacing role elimination.

For BPO hubs, the transition is complex. A 2025 IMF working paper by M. Cucio and T. Hennig found that around one-third of occupations in the Philippines are highly exposed to AI. However, 61 percent of these exposed jobs are complementary to AI, meaning technology will likely boost productivity rather than replace workers. Only 14 percent of the workforce holds jobs with low complementarity, exposing them to outright displacement.

Yet the nature of work is changing fundamentally. PwC’s 2025 Global AI Jobs Barometer found that skills for AI-exposed jobs are changing 66 percent faster than other roles, and workers possessing AI skills command a 56 percent wage premium globally.

The GCC Model as an Upgrading Pathway

To survive, service-export economies must abandon the third-party BPO model and upgrade to GCCs. Traditional BPOs operate on razor-thin margins and volume-based contracts. They bill by the hour or by the head. When AI reduces the headcount, it destroys their topline revenue model.

To survive, service-export economies must abandon the third-party BPO model and upgrade to GCCs.

GCCs are captive, wholly-owned extensions of the parent company. When JPMorgan Chase or Target builds a GCC in a low-cost site, they are not looking for cheap labour for back-office management. They are building a strategic asset.

Because GCCs handle proprietary data, core engineering, and complex compliance tasks, they are more insulated from commoditised automation shocks. More importantly, parent companies are willing to fund the reskilling. If a bank can retrain a back-office worker to audit AI compliance rather than manually enter data, both the bank and the local economy win.

Scenario Analysis and Policy Imperatives (2025–2030)

Given the pace of AI adoption, point forecasts are useless. Four divergent trajectories emerge for routine service-export labour demand by 2030, indexed to a 2025 baseline of 100:

• Passive Adaptation is the cost of inaction. Economies that keep running volume-driven BPO face an 80 percent reduction in routine labour demand by 2030. This is not theoretical. End-to-end workflow automation is already compressing task volumes at a rate that volume-billing BPO contracts cannot absorb.
• Managed Transition assumes some reskilling and GCC expansion, but no systemic policy commitment. Routine labour demand falls 40 percent. The economy survives but does not recover. This is the most likely outcome for governments that respond too late.
• GCC Moderate Offset is the floor of strategic ambition. Employer-funded reskilling, procurement reform, and targeted data governance investment offset routine job losses with higher-value roles, yielding net growth of around 5 percent by 2030. This is survival, not transformation.
• GCC Aggressive Growth is where transformation occurs. Sovereign investment in digital infrastructure, employer co-funded reskilling, and a data policy framework that positions the country as a trusted AI governance hub together yield net labour demand growth of 20 percent by 2030.

This aggressive growth is already happening. The NASSCOM-Zinnov India GCC Landscape 2026 report puts India’s GCC workforce at 2.36 million today, on track to exceed 3 million by 2030. The Philippines is targeting US$59 billion in IT-BPM revenues by 2028, per IBPAP’s January 2025 report. The gap between +5 percent and +20 percent is the quantified return on policy ambition, a core question facing every BPO-dependent economy today.

Governments and industry leaders must act fast across four imperatives:

1. Redefining Success Metrics:Governments must abandon headcount-focused incentives. Tax breaks should not target the number of desks filled. Instead, success metrics must focus on export earnings per worker, wage growth, and the share of high-complexity services delivered.
2. Investing in Governance as an Export:As frontier AI gets cheaper, governing it gets more expensive. The demand for KYC and compliance monitoring will explode. Countries establishing clear data protection frameworks can turn regulatory oversight into a lucrative export, and if paired with ease of operations, global firms will route their high-risk agentic workflows there.
3. Workforce Redesign:GCC growth will not absorb displaced BPO workers at scale; the structural employment gap is real. The priority must shift from job preservation to workforce redesign, not just upskilling. Reskilling must target super-specialty roles at the intersection of automation and compliance. Governments must act fast through employer co-funded training requirements and rapid redeployment programs for early-career workers at highest risk. The longer horizon matters too: AI literacy belongs in the primary school curriculum. But that investment pays out in 2040.
4. Infrastructure and Data Policy Readiness:AI-enabled service exports depend on reliable compute capacity, cloud access, energy infrastructure, and clear cross-border data transfer frameworks. Hardware supply chains dictate the timeline. A SemiAnalysis note in March 2026 on Nvidia’s GTC hardware roadmap highlighted that volume shipments of reworked Blackwell servers are set for late 2026 and early 2027. This delay means developing hubs have a brief, high-stakes window to build their own local compute clusters and secure energy grid allocations before the next wave of agentic systems goes live. Countries that build this complementary environment will draw the highest economic dividend. Those who do not will find themselves structurally excluded.

The real choice for service-export economies is not whether AI disrupts routine work — that is already underway — but whether policy turns that disruption into an upgrade or a retreat. GCCs, reskilling, and data governance are the levers; the window to pull them is closing. What happens next is a matter of ambition, not inevitability.

Karthik Yogeshwarleads Workforce Strategy and Operations with a focus on scaling global operations through AI and automation.

The views expressed above belong to the author(s). ORF research and analyses now available on Telegram! Click here to access our curated content — blogs, longforms and interviews.

Author

Karthik Yogeshwar

Karthik Yogeshwar leads Workforce Strategy and Operations with a focus on scaling global operations through AI and automation. He designs and optimises workforce models across …

Related Search Terms

Publications

Agentic AI and the Future of Work in Service-Export Economies

Indian Economy | Economic Reforms | Developing and Emerging Economies | Cyber and Technology | Artificial Intelligence

Jul 21, 2026

Beyond Enrolment: India’s Higher Education Completion Gap

Education in India | Skilling | Education

Jul 21, 2026