AI 正迫使印度 GCC 走出成本套利模式

India has more than 1,700 global capability centres (GCCs), but only about 5% have reached the stage where they run entire global functions from India. Artificial intelligence (AI) is widening the gap between centres that execute work and those that own products, budgets and business decisions, technology leaders said at Cypher 2026 in Bengaluru.

The distinction matters as AI makes software development and other knowledge work cheaper to execute, weakening the cost advantage that helped drive India's GCC growth. The next phase of expansion will depend on whether centres can take on greater responsibility for business outcomes rather than simply deliver work assigned by their global headquarters.

During a panel moderated by veteran journalist Karan Thapar, Rupesh Kumar, Chief Product and Technology Officer at WeWork India, Anshu Sharma Raja, Group Chief Information Officer at Arab Bank, and Ramit Bhardwaj, Chief Technology Officer at IndiQube, discussed how AI is changing the GCC model.

The panellists agreed that AI is accelerating the transition toward global delivery centres (GDCs) while exposing how far many GCCs still are from that goal.

“I would say both,” Kumar said.

Kumar divided GCCs into four maturity stages. At the bottom are centres focused on execution and cost savings, followed by those that have established efficient delivery operations. The third stage comprises centres whose local leaders own products and decisions. At the fourth, the site head operates as a CXO, running entire global functions from India.

AI exposes the limitations of the first two stages, where much of the work can be automated. At the upper levels, it can help teams deliver more, he said.

Zinnov-Nasscom's FY26 data highlights the imbalance. About 56% of India's GCCs fall within the two lowest stages, while only 5% have reached the highest.

“Stage four is difficult. We are not there yet," Kumar said.

For Raja, AI also challenges the assumption that large teams are essential to delivering value.

“With AI, you don't need a large team. You need fewer people, but sharper outcomes," she said.

She does not view this as a threat to India's technology industry, but cautioned against treating AI as another outsourcing opportunity. India needs to build research capabilities rather than depend primarily on executing work developed elsewhere.

Bhardwaj, however, sees India's cross-industry experience as an advantage. Decades of software development have given Indian technology teams domain expertise that can help them apply AI to business problems.

AI, he said, is “an efficiency multiplier.”

AI may accelerate GCCs' progress, but it cannot determine how much authority their parent companies are willing to delegate.

Bhardwaj said companies typically begin by moving work to India to reduce costs. As confidence builds, they transfer more complex responsibilities, eventually allowing teams to manage the full product lifecycle. AI could speed up this progression by making code cheaper to generate.

The final stage, however, depends on whether decision-making authority moves to India.

“It's not just about the delivery, but also about decision-making,” he said, referring to the distinction between a global capability centre and a global delivery centre.

Bhardwaj was less convinced that headquarters would readily surrender control over strategic decisions.

Kumar sees the transition accelerating. Site leaders who own products and decisions typically have profit-and-loss responsibility, he said. Centres that once took five to 10 years to reach this stage can now get there in three to four years.

Raja offered a practical definition of ownership in banking: setting the technology roadmap, controlling budgets and teams, and taking responsibility when critical systems fail.

In payments, that can mean answering a call at 3 am when a transaction system breaks down. Ownership, in other words, involves accountability for the outcome, not merely responsibility for completing the work.

The shift is also visible beyond individual GCCs. IBM Consulting's Santhosh Rao, partner and executive director for India and South Asia, has said the centre of gravity for decision-making is shifting. Global financial firms are also bringing strategic work in-house that IT service providers previously handled.

Another estimate puts the share of GCCs that own global products and platforms at 44%, suggesting that ownership of specific products and platforms is more widespread than full control over global functions. The two measures, however, reflect different levels of responsibility.

The shift towards smaller, more productive teams has yet to translate into weaker demand for office space, according to Kumar and Bhardwaj, whose companies lease workspaces to GCCs.

Kumar said clients increasingly prioritise quality, speed and flexibility as workforce requirements become harder to predict. GCCs account for about 40% of WeWork India's revenue this year, and clients expect steady growth over the next three to four years.

AI is also enabling centres to solve problems and use data they previously couldn't, he said, creating new reasons to expand operations.

A report by research firm RedSeer, prepared with workspace operators WeWork India and SmartWorks, projects India's GCC AI workforce to grow from about 1.8 lakh to 7.3 lakh by 2030.

Bhardwaj said the number of seats required by individual clients is moderating as AI enables smaller teams to do more. But the number of GCCs continues to grow.

“We are not actually witnessing any slowdown,” he said.

Raja, speaking from the tenant's perspective, said rising rents and demand for new office buildings remain concerns. In some cases, buildings fill up before leases are finalised.

The picture is therefore more complicated than a simple reduction in headcount. AI could reduce the number of employees needed for a given function while increasing the range of work a centre handles, allowing the broader GCC ecosystem to expand even as individual teams become leaner.

Ownership of IP and the Need for a Mindset Shift

The question of who owns the intellectual property (IP) created by GCCs also came up during the discussion.

Bhardwaj and Raja said that IP developed within a GCC generally belongs to its parent company. Kumar argued that ownership is closely tied to who defines the problem being solved, which is usually the global headquarters.

He also predicted that governments would seek greater control over foundational AI models.

“No government will allow the AI to go out of their control,” he said.

An audience member raised the growing use of forward deployed engineers (FDEs), who work closely with customers to build and deploy AI systems. Bhardwaj described the model as a repackaging of work that GCC teams have long performed.

The model is gaining traction across the industry. Big Tech companies are investing billions in FDE teams, while Indian IT services firms such as TCS are adopting similar approaches.

The discussion also turned to whether Indian companies are willing to make the investments needed to build globally competitive technology products. Kumar said many large Indian companies continue to operate as free-cash-flow generators rather than take the risks associated with ambitious technology bets.

He cited former Infosys CEO Vishal Sikka's reported push for a larger investment in OpenAI, which Infosys did not pursue.

For Raja, the underlying problem is how Indian companies see themselves.

Many still identify primarily as IT services organisations, she said, rather than technology companies capable of building products and owning strategic decisions.

“We literally have to shed that skin,” she said.