‘Stop measuring AI by ROI, focus on return on experimentation’: NTT DATA’s Sudhir Chaturvedi

Japanese technology services and digital infrastructure major NTT DATA is betting that the next phase of enterprise AI will be driven as much by infrastructure and services as by frontier models. Part of the NTT Group, one of the world’s largest telecommunications companies with annual revenue of about $93 billion, NTT DATA generates over $30 billion in annual revenue, serves 75% of the Fortune Global 100, and is the world’s third-largest data centre provider. In India, where it is the largest data centre provider, the company has invested more than $3 billion since 2011 and has announced another $1.5 billion investment over the next three years, alongside continued investments in AI and digital infrastructure.

In an interaction with Fortune India, Sudhir Chaturvedi, Global Chief Growth Officer and CEO, North America, NTT DATA, argues that enterprises are asking the wrong question when evaluating AI. Instead of focusing on return on investment, companies should first measure “return on experimentation”, he says, while outlining why services, responsible AI and infrastructure will define the next stage of enterprise adoption.

Enterprises are spending billions on AI, but many still struggle to demonstrate returns.

I think ROI is the wrong metric to look at right now. We’re dealing with breakthrough technology that is still going through its maturity phase. The models are evolving, pricing models are changing and even the commercial frameworks around AI are still being developed.

Instead of focusing on return on investment, enterprises should think about return on experimentation (ROE).

Companies shouldn’t experiment everywhere. They should identify two or three business functions where AI can make the biggest difference—whether that’s supply chain, underwriting, lending or financial crime—and go really deep.

Without experimentation, you don’t understand what’s actually possible with the technology. Once organisations figure out how AI fits into their business, the ROI will naturally follow. The second aspect people often overlook is personal productivity. Every meeting today is better prepared because people are using AI. The quality of interactions has improved significantly across organisations. That’s difficult to measure immediately in financial statements, but it’s real.

So where are enterprises actually seeing value today if not traditional ROI?

Productivity is one area, but we’re also seeing organisations become much better at solving specific business problems. Take supply chains. Given today’s geopolitical environment, planning has become much more complex. AI is extremely good at scenario planning because it can analyse multiple data points simultaneously and help decision-makers evaluate different possibilities.

Similarly, in insurance we’re focusing on claims and underwriting, while in banking we’re looking at lending and financial crime. The important thing is to experiment in areas where there’s already structured data and well-understood business processes. That’s where AI can create meaningful improvements.

The conversation has now shifted towards autonomous AI agents. At the same time, recent incidents have raised concerns around safety and cybersecurity. How should enterprises approach this?

Enterprises should treat this as both a challenge and an opportunity. From a cybersecurity perspective, these incidents highlight the importance of strengthening an organisation’s overall security posture.

At NTT DATA, we’ve consistently advocated responsible AI. Before deploying autonomous agents, organisations need governance frameworks, access controls, decision rights and continuous monitoring.

Our philosophy is human at the core, not just human in the loop. Agents shouldn’t have unrestricted autonomy. They should operate within clearly defined guardrails and escalate to humans whenever they encounter situations beyond their context.

Millions of enterprise AI agents are already operating successfully today. We ourselves have deployed around a hundred agents across our infrastructure services business, supporting order management, software support and operational workflows. They work because they’re designed with governance from the outset.

Open-source models have become increasingly capable. What does this mean for enterprises who are using frontier models like Claude and ChatGPT?

Yes. Most enterprise customers don’t want to depend on a single model provider. They’re building architectures that allow them to choose whichever model is most appropriate for a particular workload.

Not every use case requires the latest frontier model. Some workloads can be handled perfectly well by smaller or open-source models.

The responsibility of services companies like ours is to help customers choose the right model for the right task, orchestrate multiple models together and ensure they operate within appropriate governance frameworks.

Frontier AI companies are increasingly expanding into enterprise services. Do you see them becoming competitors?

I actually think they’ve realised how complex enterprise environments really are. Building a model is one thing. Understanding how a mortgage business works, how insurance claims are processed or how global supply chains operate is something entirely different. Enterprise complexity is where competitive advantage exists. Every company performs similar functions, but the way they execute those processes is what differentiates them.

That’s why I often say the future of AI won’t be won by models alone. It will be won through services. Interestingly, many of the frontier AI companies themselves are now building services capabilities because they’ve realised enterprise deployment requires much deeper business understanding.

AI is also reshaping hiring across the IT industry. What does that mean for India’s workforce?

Technology roles have never remained static. Over the years we’ve moved from specialised programmers to full-stack engineers. AI represents another transition. Some tasks like coding and testing will increasingly be automated, but entirely new responsibilities are emerging around AI governance, data foundations, agent orchestration and business process transformation. Domain expertise will become far more valuable because AI models still don’t understand business processes the way humans do.

We’re continuing to hire. In India alone, we’ll add more than 5,000 people this year. The roles will evolve, but the opportunities will continue to exist. We’re also investing heavily in reskilling. Every leader in our organisation is expected to undergo AI training because this transition has to begin from the top.

NTT DATA describes itself as a ‘full-stack transformation company’. What does that mean, and why has that become more relevant in the AI era?

Full stack means we operate across every layer of enterprise technology—from global networks, subsea cables and data centres to cloud, cybersecurity, applications and business consulting.

That’s unusual in today’s market. Many competitors have moved towards asset-light models, especially after the shift to cloud computing. But AI has made infrastructure strategically important again.

Today, enterprises need both AI infrastructure and applications. Because we operate across the entire technology stack, we’re seeing growth across every part of the business. Infrastructure investments, particularly around AI, will continue for several years, and eventually applications will build on top of that foundation.

Does NTT Group’s telecom heritage give NTT DATA an advantage as infrastructure becomes central to AI?

Yes, because AI requires significantly more infrastructure than previous technology waves. Cloud infrastructure will continue to grow, but AI is creating entirely new demand.

Many enterprises are also looking at private AI infrastructure for reasons such as sovereignty and intellectual property protection. Today’s CIOs are effectively managing two environments simultaneously—they have to run existing systems while building entirely new AI-enabled businesses. That’s where infrastructure has once again become a strategic differentiator.

NTT DATA recently announced another $1.5 billion investment in India. Where will that investment be directed?

A significant portion will continue to support our digital infrastructure business. India is already our largest data centre market, and we’re investing in GPU infrastructure and working with hyperscalers to provide AI capacity.

Beyond infrastructure, we also see India becoming an increasingly important enterprise AI market. The opportunity is no longer limited to traditional IT services. AI allows us to participate directly in transforming business operations, which is a much larger market.