Volume 4: An Exclusive Feature on Why AI Needs a New Operating Model

Labor arbitrage is over for UK enterprises. How domain intelligence and agentic AI are replacing headcount-based models and what to do about it now.

Labor arbitrage is over for UK enterprises. How domain intelligence and agentic AI are replacing headcount-based models and what to do about it now.
May 20, 2026
Venugopala Reddy Dumpala
Solutions Head, UK & Middle East, Firstsource
Download now
Labor arbitrage is over for UK enterprises. How domain intelligence and agentic AI are replacing headcount-based models and what to do about it now.

For nearly three decades, the outsourcing playbook ran on a single logic: find labor-intensive processes, find geographies where that labor costs less, and arbitrage the difference. In banking, telecommunications, energy, and retail, the model scaled reliably. It built entire businesses.

The UK of 2026 looks nothing like the UK that built that model. Brexit restructured trade and talent flows. Inflation cumulated at over 20% between 2021 and 2024.¹ Interest rates were held at 15-year highs before they began to ease.² Between 2016 and 2024, the UK lost an estimated 785,000 EU workers from its labor force.³ In financial services alone, thousands of roles migrated to Dublin, Frankfurt, and Amsterdam. Tech salaries rose nearly 10% year-on-year as digital demand outpaced supply.⁴

At the same time, agentic AI is rewriting the economics of service delivery. These forces are not arriving sequentially. They are converging, and they are converging now.

The question facing UK enterprises has changed. It is no longer "Where can I find cheaper labor?" It is "Where can I find the core intelligence: data, operational and process intelligence that helps us deliver better outcomes for our customers, and how do we deploy it in a way that builds capability rather than hollows it out?"

Over the past two years, I have observed a significant shift in how enterprises operate, with the pace of change accelerating rapidly. A systemically large UK enterprise that focused on labor arbitrage for over two decades is now rapidly repositioning itself as a flag bearer for AI across its operations.

Why labor arbitrage has run its course

The Global Capability Center (GCC) model was a rational response to the pressures of the last decade. Today, over 1,800 GCCs operate out of India alone, employing nearly 1.9 million professionals, growing at 11 to 12% annually.⁵ The logic was sound: retain process control, access lower-cost talent.

But the GCC model carries forward a fundamental assumption: that labor cost is the primary lever of value creation. Work is defined in FTEs; value is measured in cost-per-transaction, and the competitive moat is, ultimately, the wage differential between London and Bangalore. That moat is eroding from both ends.

Indian IT services wages have risen 8 to 10% annually,⁶ compressing the offshore advantage. More consequentially, agentic AI is decoupling the relationship between headcount and output entirely. When an intelligent agent can process a payment or mortgage application, triage a customer complaint, or reconcile a regulatory report in seconds, the cost of the human operator becomes secondary to the quality of the intelligence applied.

If your operating model still defines value in FTEs and cost-per-transaction, the ground has already shifted beneath it. The question is not when to respond, but how far behind the response has fallen.

Three shifts redefining how UK enterprises compete

Intelligence That Operates is Firstsource's response to exactly this shift: AI that doesn't just advise or automate, but actively runs complex operations end-to-end, with domain expertise encoded into it and accountability built around it.

Across the conversations I have with UK enterprises, three shifts keep surfacing that point in exactly that direction. They are not predictions. They are the ground shifting beneath every operating model in the UK right now.

The first is that domain intelligence has become the competitive differentiator that labor cost used to be. In regulated industries, and in the UK that means almost every industry, deep operational expertise is no longer a background requirement. The FCA's Consumer Duty, in full force since 2024, changed the accountability framework for financial institutions fundamentally. Compliance is no longer delegable. Firms must demonstrate that customer outcomes are embedded in every operational decision.

Ofcom's evolving posture on digital markets imposes similar rigor on telecommunications providers. Ofgem's expansion of regulatory oversight to 14,000+ heat network operators⁷ reflects the growing complexity of the energy transition. What this means in practice is that generic AI capability, deployed without that accumulated domain intelligence behind it, cannot meet the bar these frameworks set. The enterprises pulling ahead are those that have encoded years of sector-specific understanding into how their AI systems reason and decide.

Intelligence arbitrage, not labor arbitrage, is the moat that is forming.

The second shift concerns what agentic AI must do to be worth the investment. Only 5 to 12% of UK enterprises have deployed genuine agentic systems: AI that can reason, decide, and act within defined operational guardrails.⁸ The market for agentic AI is projected to grow from approximately $7.5 billion in 2024 to nearly $200 billion by 2034.⁹ But the speed of deployment is not differentiating.

The differentiating question is whether intelligence compounds. Agentic systems that simply execute a defined task at pace are, increasingly, table stakes. What creates a durable advantage is AI that gets measurably sharper with every engagement: learning from exceptions, refining decisions, building an accumulated edge that a competitor deploying the same model six months later cannot simply replicate.

A bank whose anti-money laundering system has processed five years of domain-specific decisions, and learned from every one of them, is operating in a different category from a bank that has deployed the same underlying model without that depth. Speed is replicable. Accumulated intelligence is not. This is what we see as the most sophisticated UK enterprises beginning to orient around, and it is closely aligned with how we think about Intelligence That Operates.

The third shift is the one that receives the least attention and may carry the most long-term consequences. When processes move offshore, the knowledge embedded in those processes moves with them. Over time, the institutional understanding of how to adjudicate a mortgage loan, resolve a complex billing dispute, or navigate energy switching regulations becomes resident in Mumbai, Bengaluru, Hyderabad, or Manila rather than Manchester or Birmingham.

Against a backdrop of 257,000 British nationals emigrating in 2024,¹⁰ and surveys suggesting 38% of professionals aged 25 to 34 are considering leaving,¹¹ this is not a labor market footnote. It is a strategic vulnerability that most UK enterprises have not yet fully priced in, and it is arriving now. Domain expertise is becoming the primary source of competitive advantage.

The UK enterprises responding well to this are building AI-native operating models around domestic regulatory and commercial context, creating roles that are onshore by design: domain architects, AI trainers, regulatory intelligence specialists, and outcome designers. The offshore model hollowed out institutional knowledge. The intelligence model, governed well, can rebuild it.

The pattern across verticals

The application looks different by sector, but the underlying logic is consistent. In each case, the question is the same: how do you encode deep operational intelligence into AI systems that can act on it, and hold the outcome accountable end-to-end?

In banking and financial services, that means building AI-native compliance frameworks that turn Consumer Duty from a cost center into a competitive differentiator, and intelligent collections systems that balance recovery rates with genuine customer outcomes.

In communications and technology, it translates to moving beyond agentic systems that diagnose network issues and proactively resolve billing queries, not because a script directs them, but because they understand the customer's context.

In energy and utilities, it means having the operational depth to navigate the net-zero transition with intelligent operations capable of managing smart grid integration, demand response, and multi-fuel billing simultaneously, while maintaining compliance with a regulatory framework still being written.

In retail and e-commerce, it means building supply chain intelligence that adapts in real time, customer engagement that is truly personalized, and returns management that balances satisfaction with margin.

Across every vertical, the enterprises that will thrive are those that stop optimizing cost and start optimizing intelligence.

From headwinds to advantage: what UK enterprises should do now

The UK retains an advantage that does not always appear in the economic data: world-class innovation infrastructure, a financial services ecosystem shaped by decades of regulatory sophistication, and an institutional appetite for technology that few markets can match.

DeepMind was built here. The fintech revolution was substantially a London story. That foundation matters, because the three shifts described previously are not abstract global forces. They are playing here, in sectors the UK already understands deeply, and that gives UK enterprises a head start that should be acted on.

The economic headwinds of the past decade have been painful. They have also been clarifying. They revealed which operating models are resilient and which are merely cheap. They showed that regulatory complexity is not a burden to offload but a domain to master.

What I tell clients who ask where to start is this, and these are not theoretical steps. They are the three things I have seen that make the most difference in the conversations I have had over the last 18 months.

First, audit your AI ambition against your AI reality. If agentic deployment remains in pilot, the gap between aspiration and production is a competitive liability.

Second, audit every process you are considering offloading against one question: is intelligence embedded in that process something you can afford to lose? If the answer is no, the decision is not whether to offload it, but who you trust to carry that intelligence forward, encode it into how your AI systems operate, and remain accountable for what comes out the other side.

Third, look for partners who hold accountability for outcomes, not engagements. The partner who designs the transformation, implements the architecture, and manages it in production, with full accountability for what it delivers, is a different relationship from a vendor who hands off at go-live.

The labor arbitrage chapter is closed. The enterprises that recognize this earliest will not just cut costs differently. They will compound intelligence in a way that widens the gap between themselves and those still optimizing headcount.

References:

Download UnBPOTM Quarterly