From Cost Hub to Intelligence Engine: Why the Traditional GCC Value Equation Is Broken

The math behind the Global Capability Center was never complicated. Take the cost of a role in a high-cost market, subtract the cost of the same role in a lower-cost location, multiply by headcount, and the difference is the business case. That equation funded the GCC industry for thirty years and still shows up in board decks today.

It no longer holds up on its own, and the data on why is now hard to ignore.

The original GCC value proposition rested on a single variable: the location differential. Centralize enough roles in a lower-cost market, standardize the process around them, and the savings scale with headcount. It is a labor-arbitrage model, and as a pure cost play, it worked for a long time.

The problem is that every input to that equation is now moving in the wrong direction at once.

The premium for the talent that matters most is rising faster than the differential it was supposed to protect.

McKinsey Global Institute research found that demand for AI fluency, the ability to use and manage AI tools, has grown sevenfold in two years, faster than for any other skill category in the job market, and that global demand for applied AI talent more than tripled between 2018 and 2025. That growth is concentrated in exactly the skills a GCC now needs most: engineers who can build, govern, and supervise agentic systems. As that talent gets scarcer everywhere, the cost gap the original equation depended on gets narrower.

Lift-and-shift savings plateau, and most centers are still built on lift-and-shift.

McKinsey’s State of Organizations 2026 research is direct about this: traditional centralization models bring near-term labor cost savings, but those savings plateau quickly once the initial migration is complete. There is no second wave of savings built into a model whose only lever is headcount location.

The center is still measured on the wrong output. According to the same research, shared-services and capability centers today spend the bulk of their effort on operational tasks (56 percent of centers) and administrative tasks (49 percent), with far fewer focused on analytical work (41 percent) and only 26 percent on strategic or advisory work. A center optimized for transactional throughput will keep reporting transactional metrics, cost per ticket, SLA adherence, hours logged, none of which capture whether the center is creating enterprise value.

Layer geopolitical exposure on top of this and the cost-arbitrage case weakens further. Almost three in four leaders (72 percent) in McKinsey’s survey say geopolitical uncertainty has had a notable impact on their organization, and centers built around a single low-cost location carry that risk directly. Technology, not geography, is now what gives organizations the flexibility to route work around disruption rather than being exposed to it.

The variable that used to drive the value equation was location. The variable that drives it now is intelligence: how much AI-augmented engineering capacity a center can generate and orchestrate, independent of where the people sit.

This is the logic behind what Ascendion calls AI Arbitrage. It is not a rebrand of labor arbitrage. It is a different source of value entirely, and McKinsey’s research puts a number on the gap between the two models. Centers that shift from a traditional shared-services structure to an AI-native operating model see an additional 20 percent gain in cost effectiveness beyond location savings alone, a 40-fold increase in access to innovation, and roughly a 50 percent productivity improvement tied to operational excellence. Customer and user experience metrics improve by another 20 percent on top of that. None of those gains come from moving the work somewhere cheaper. They come from redesigning what the work is.

That redesign also changes what the center actually delivers. A cost hub sells hours. An intelligence engine delivers Services-as-Software: governed, reusable AI-driven capability that scales with demand instead of headcount, and that gets more valuable the more it is used rather than more expensive.

The clearest sign a center is still running on the old equation is what shows up in its monthly business review. Headcount ratios, cost per FTE, and SLA compliance measure whether the center is efficient at the work it was given. They say nothing about whether the center is generating outcomes: faster cycle times, fewer defects, shorter time to market, better resolution rates.

Enterprises that make the shift do not add outcome metrics alongside the old ones. They replace them. The center’s mandate moves from “deliver the agreed volume of work at the agreed cost” to “own the outcome, and use whatever mix of AI agents and specialists gets there fastest.” That is a governance change before it is a technology change, and it is usually the one enterprises skip.

None of this is theoretical resistance to change. McKinsey’s research shows most enterprises are already trying to move and getting stuck. Eighty-four percent of leaders plan to expand the scope of their shared-services and capability centers in the next one to two years, but more than 40 percent have not yet started systematically adopting the technologies that expansion requires, and only 6 percent of business services leaders report realizing full value from advanced technologies across multiple use cases.

The two most cited barriers are integration with legacy systems (42 percent) and organizational resistance to change (41 percent). Both point to the same root cause: enterprises are trying to bolt AI onto a center that was architected around the old equation, rather than re-architecting the center around AI. Integration friction is what happens when agentic workflows are layered onto systems and governance built for headcount management. Resistance is what happens when a center’s incentives, KPIs, and career paths are still built around the model being replaced.

Rebuilding the value equation is a workflow and governance problem before it is a tooling problem. It requires redesigning end-to-end processes around AI rather than automating pieces of the existing process, building a governed platform layer so agentic capability compounds instead of fragmenting across one-off pilots, and moving the center’s scorecard from throughput to outcomes. This is the operating model behind Ascendion’s approach to Engineering to the Power of AI: a Carbon + Silicon model where AI agents and engineers are deliberately orchestrated around a shared outcome, not stacked on top of each other.