At the Bengaluru GCC Summit this year, the mood was triumphant. India’s Global Capability Centres, once dismissed as glorified back offices, are now routinely described as the innovation engines of the world’s largest companies. The numbers back the swagger: 2,117 GCCs now operate in India, employing 2.36 million people and generating $98.4 billion in revenue, according to the Zinnov-Nasscom India GCC Landscape 2026 report. Ninety-six percent of centres launched since FY21 began life with a product, engineering, or R&D mandate rather than a cost-arbitrage brief. Zinnov’s most striking finding is that nearly half of India’s GCCs now do work comparable in sophistication to what happens at headquarters.
Buried in a less celebratory piece of research, though, is a number that should worry every CEO who has staked part of their operating model on this transformation. A 2026 analysis by the GCC advisory firm 10decoders found that 83% of India-based centres are now considered mission-critical to their parent organisation. Only 17% hold genuine ownership over the strategic decisions and budgets tied to that work. The gap between those two figures is not a rounding error. It is the real story of the GCC boom, and it is one that most of the “cost centre to innovation hub” narrative conveniently skips.
Indispensable, but not in charge
What “mission-critical but not in control” looks like in practice is an engineering team in Hyderabad architecting the agentic AI layer for a global logistics platform, running the models that decide routing and pricing in real time, while final sign-off on the product roadmap sits with a director in Chicago who sees the work in a monthly steering deck. It is a data science group in Bengaluru that has quietly become the best in the company at a particular class of problem, yet still routes every material decision through a headquarters that increasingly lacks the day-to-day context to evaluate it well.
The HFS Generative GCC Index puts a number on this maturation gap: execution-only centres score 45.2 out of 100 on its maturity scale, while centres that have progressed to owning a product end-to-end score 70.6. India’s ecosystem average sits at 61.9, a midpoint that reflects an industry caught between two operating models rather than one that has decisively chosen the second. Only around 27% of centres reach what the index calls “portfolio hub” status, full ownership of a product or platform, within five years of launch. Everyone else remains somewhere on a spectrum between “does the work” and “decides what the work should be.”
This is not simply an HR problem, though it shows up first in HR metrics. Attrition in mid-level technology roles in Bengaluru and Hyderabad runs at 18 to 25% annually, with mid-to-senior data and technology roles seeing turnover as high as 30%, well above what the industry considers healthy. A recurring theme in conversations with GCC leaders, echoed in a recent Intelligent CIO analysis of the sector, is that ambitious engineers “lose their spark” when they are permanently boxed into execution rather than given a stake in direction. Companies have responded, tellingly, not by redesigning authority but by relocating the arbitrage: tier-two cities such as Coimbatore, Kochi, and Ahmedabad now show attrition rates 10 to 15 percentage points lower than the metros, and account for roughly 12% of new GCC hiring. That is a rational cost-management move. It is also, if the underlying grievance is about ownership rather than salary, a way of buying a few more years before the same ceiling reappears somewhere else.
Why headquarters holds on
It would be convenient to treat the authority gap as an oversight, something that will close naturally as centres mature. It is worth taking seriously the alternative explanation: that many multinationals are withholding decision rights deliberately, because devolving them raises questions boards would rather not confront yet. If the team doing the most sophisticated product thinking sits 12,000 kilometres from the corner office, what does that imply about where the company’s real centre of technical gravity now sits, about succession planning for senior technical roles, about which office should have a seat closer to the board table? It is far easier to keep the GCC “mission-critical,” a phrase that flatters the centre while changing nothing about who signs off, than to answer those questions directly.
That reluctance was manageable when GCCs were still mostly extensions of the delivery organisation. It is becoming harder to defend now that GCCs sit on the sharpest edge of enterprise AI work. Zinnov counts more than 1,200 India GCCs with active AI and machine learning capability, and 52% of AI use cases inside these centres now involve generative or agentic systems, overtaking traditional automation for the first time. India employs roughly 550,000 AI and machine learning professionals, about half of them inside GCCs, making the country the largest single pool of applied agentic AI talent tied to any group of multinational employers anywhere. Routing judgment calls about that work through a headquarters that did not build the systems, and often cannot fully audit them, is not a governance nicety. It is a latency and quality problem dressed up as prudence.
The visa shock that changes the calculus
Geopolitics has just made this harder to defer. Washington’s increase of the H-1B visa fee to $100,000, which took effect in late 2025, was significant enough that JPMorgan economists estimated it could remove roughly 5,500 work-visa moves a month from the system, turning what used to be a routine relocation decision into what the bank’s analysts called “a high-stakes strategic decision.” The practical effect for multinationals has been to accelerate the shift of decision-adjacent technical work to India rather than to the United States, simply because moving the person is now prohibitively expensive while moving the work is not. Engineering R&D-heavy centres are growing 1.3 times faster than traditional GCCs as a result, and banking alone now runs more than 90 India centres across over 50 institutions, employing more than 180,000 people in roles that increasingly touch risk, compliance, and product decisions once reserved for head office.
This is the trap boards need to see clearly. The visa shock increases the volume and sensitivity of work flowing to India. It does nothing to increase the authority that comes with it. Companies can now find themselves with more mission-critical, higher-stakes work sitting in centres that still operate under the old 17% ownership model, widening exactly the gap that is already driving attrition and, increasingly, quiet underperformance in decision speed.
What agentic AI is about to do to this arrangement
The least discussed implication is also the sharpest one. Agentic AI is best at replacing well-specified, replicable, delivery-oriented work, precisely the profile of an execution-only GCC scoring near 45 on the maturity index. It is much worse at replacing the tacit judgment, contextual authority, and end-to-end accountability that sits inside a genuine portfolio hub. Over the next several years, that difference is likely to sort India’s GCC population into two very different futures. Centres that were given real ownership will compound their advantage, because the thing that makes them hard to automate is exactly the decision-making authority their parent company was reluctant to grant. Centres that were kept in a purely executional role, however sophisticated the code they write, are the ones most exposed to being quietly absorbed by the same agentic systems they helped build.
Put bluntly, the multinationals that spent a decade being careful not to give their India centres too much authority may have inadvertently placed those same centres in the automation blast radius. The ones that took the harder political step of devolving real ownership will find that decision now looks less like generosity and more like foresight.
For CEOs and boards, the sensible response is not another rebrand from “delivery centre” to “innovation hub,” a relabeling exercise India’s GCC sector has already been through twice. It is an honest audit of where decision rights actually sit relative to where the best technical judgment in the company now resides, followed by a deliberate redesign, product ownership, budget authority, and board-level visibility extended to the centres whose maturity already justifies it. The GCC sector did not create this reckoning. It simply grew fast enough, and got good enough, to make the question impossible to keep avoiding.
