The fear is that AI tools capable of automating coding, legal review, sales workflows and other enterprise functions could erode the high-margin application services revenues that account for 40–70% of Indian IT companies’ business.
Jefferies has warned of more pain ahead, arguing that AI-led automation could structurally disrupt application services. Motilal Oswal estimates that close to 10% of industry revenues may be vulnerable. HSBC has said the risks this time may be more credible than in previous technology cycles. JP Morgan has cautioned that companies may find it difficult to shield themselves from the next wave of automation.
Yet, in a conversation with CNBC-TV18, industry veterans Mohandas Pai, Chairman of Aarin Capital and former CFO of Infosys, and R. Chandrasekhar, former President of NASSCOM and former Secretary of the IT Department, pushed back against the notion that February 4 marked a day of reckoning for Indian IT.
Pai argued that the narrative around an existential threat is overstated. He pointed to the sheer scale of legacy software embedded in global enterprises — an estimated $20–25 trillion worth across the US and Europe — with annual technology spending of $1.2–1.75 trillion. That installed base, he suggested, cannot be replaced or automated away overnight by plugging in a new AI tool.
He emphasised that AI tools such as Claude can assist in writing portions of new code but are far less effective when dealing with complex legacy systems. Most global enterprises continue to use AI selectively, particularly in research, discovery and limited productivity use cases. Comprehensive AI-led transformation of core operations remains at an early stage.
The more nuanced question, therefore, is not whether AI will have an impact, but how much of Indian IT revenues are truly exposed.
Pai estimated that tools like Claude may directly apply to roughly 25–30% of total work in application development and related services. While application services form a significant share of revenues for large IT firms, the nature of work within that bucket varies widely. Large-scale system integration, legacy modernisation, database consolidation, testing, governance and compliance layers are not easily automated.
Chandrasekhar echoed the view that while AI will significantly improve productivity, it does not eliminate the need for technology partners. He noted that every major technological shift — from Y2K remediation to the rise of platforms and SaaS — has been accompanied by predictions of industry collapse. AI, in his view, represents a powerful productivity accelerator, but not an overnight structural wipe-out.
That said, both acknowledged that the traditional labour-arbitrage model is under strain.
The Indian IT services business historically scaled revenues through large teams and billable hours. AI-assisted development is already delivering productivity gains of 8–15%, with expectations of 15–30% improvements over the next 12–18 months in new code development. Developers who effectively use AI tools are significantly more productive than those who do not.
This implies a reduction in employment intensity. Fewer people may be required for the same output. The mix of skills will shift upward, with greater emphasis on AI fluency, architecture, governance and domain expertise. Fixed-price, output-based contracts — already accounting for 45–50% of revenues at many firms — could rise further, reducing the centrality of pure time-and-material billing.
However, Pai argued that productivity gains do not automatically translate into revenue destruction. Global enterprises, particularly large banks and manufacturers, continue to operate with fragmented systems and multiple legacy databases. Many lack unified data architectures necessary to run AI effectively. The backlog of modernisation work remains substantial.
If AI releases capital through efficiency gains, the critical question is how enterprises redeploy those savings. Pai’s view is that competitive pressure from digital-native firms will force incumbents to reinvest in technology transformation rather than simply expand margins. In that scenario, IT services firms could see more projects, even if individual project sizes shrink and execution cycles compress.
Chandrasekhar added that fear of being left behind is accelerating AI adoption across sectors such as manufacturing and healthcare. This, paradoxically, may increase dependence on established technology partners who can manage complex transitions at scale. The differentiation among IT firms will likely widen, depending on how effectively they position themselves in AI-led transformation.
Another dimension of the current debate relates to valuation narratives.
Pai suggested that new AI firms seeking multi-billion-dollar fundraises have strong incentives to frame their products as industry-disrupting breakthroughs. In an environment where earnings may not justify lofty valuations, projecting dramatic transformation can amplify investor enthusiasm. This, he implied, may contribute to the intensity of the “SaaSpocalypse” rhetoric.
Markets, however, must distinguish between narrative amplification and operational reality. The existence of advanced AI coding tools does not equate to instant enterprise-wide automation. Integrating AI into mission-critical systems requires testing, governance, data restructuring and change management — areas where service providers retain relevance.
The transition will nonetheless be disruptive. Employment growth in the sector is likely to moderate. Skill requirements will evolve rapidly. Margin structures may shift as automation reduces effort intensity. The headcount-driven expansion model that defined the past two decades is fading.
Also Read | AI fears ‘overblown’, no structural decline in IT: LTIMindtree CEO
But the broader revenue pool — anchored in trillions of dollars of accumulated software and annual technology budgets exceeding a trillion dollars — suggests that transformation, rather than extinction, is the more probable trajectory.
For investors, the debate may ultimately hinge less on whether AI will change Indian IT — it undoubtedly will — and more on which companies adapt fastest to a model where capability, not capacity, defines competitive advantage.
Below are the excerpts of the interview.
Q: February 4 — many are calling it a day of reckoning for the Indian IT industry. That’s the day when Anthropic launched its new plugins, which could automate work across legal, sales, and marketing functions. The fear is — and as Jefferies is calling it — a “SaaSpocalypse.” JP Morgan says there is no way to hide from this latest new wave of automation. Do you share that assessment — that this is an existential risk for the IT industry, and not just a simple tech cycle that we’re seeing?
Pai: I don’t think it’s an existential risk. We need to understand this better. Let me give you some data. Whenever a new innovator comes to the table, they create fear to increase their valuation. Anthropic is raising $25–30 billion and wants a high valuation. Unless they release something dramatic and say it is the end of the world and that they are the new heroes, they are not going to get that valuation, because earnings do not justify it. They have to make noise. Palantir is highly overvalued. They have to make noise.
But let me give you some data. There is between $20–25 trillion of accumulated software used by global enterprises in Europe and America. Annual spending across these major markets is between $1.2–1.75 trillion, and this includes new code being written. All this is not going to disappear in one or two days. There is no way you can simply plug something in and say, “Operate my system.” It is not going to work. And Claude is only for writing code, and even then, only part of the code.
Global enterprises are still not using AI in their core operations. They are using it for discovery and other areas. In research and development and in discovery, it works well.
Global enterprises have not yet started using AI comprehensively across their operations. To improve productivity using AI, they will need help. Who is going to help them? IT service companies. IT service companies will help global enterprises use AI in their operations to become more productive and efficient.
It is also important to understand that AI does not work effectively with legacy code. It works better with new code. Let me summarise. There is a lot of work that can only be done by service companies to help global enterprises use AI, and that will continue. When you write new code, there will be productivity gains of 15–30% over the next year to year-and-a-half. Already, 8–15% productivity gains are being delivered to clients and are in operation.
The application areas where tools like Claude can be directly used may account for 25–30% of the total work. The rest is different in nature. All IT service companies are using tools like Claude and others. They are going to see improvements and will release capital. The big question is: when capital is released, will global enterprises use it to rejuvenate technology or simply add it to the bottom line? My view is that they will use it to modernise technology and create more work with the same money so that they become more competitive. The real threat to everyone is new enterprises using the latest technology. So these companies are on top of it. There will be disruption and significant productivity changes, but the fears are overblown. Service companies speak to clients every day. They know what is happening and are working with them.
Q: Mr. Pai’s assessment is that fears are overblown and that IT companies are nowhere near losing their relevance. Let me put that to Mr. Chandrasekhar. After February 4 and the launch of Anthropic’s plugins, many are asking: if a US enterprise can automate legal review or vetting using Claude Co-Work or OpenAI Codex, why would it need a 50-member team sitting in Bengaluru?
Chandrasekhar: I think the real question is whether something so dramatic happened on February 4 that it has turned the entire IT industry upside down. I agree that the IT industry has seen many major changes before, whether it was Y2K, platforms like GitHub, or now AI. AI has been in use for quite some time. It was not born on February 4, at least as far as IT developers are concerned.
Every time there is a significant development, media hype and market hype amplify it tremendously. In the midst of that, we need to step back and take a calibrated and balanced view. Agentic AI and AI-powered platforms are not entirely new. They have been around for a while.
Am I saying this will not have a significant impact? No. Every major step in productivity improvement has implications, both for revenues and, more importantly, for manpower. The old models that depended heavily on headcount are fading away. AI has made developers who use these tools far more productive than those who do not — possibly by an order of magnitude. We are not talking about marginal gains.
So employment intensity in the IT industry is likely to reduce significantly, while quality and capability intensity will rise. The bigger question is the impact on revenues. For the reasons Mr. Pai mentioned, revenue growth is likely to continue. The availability of AI tools does not mean client companies will suddenly say they can manage everything themselves. They still need help.
In fact, the fear of being left behind is prompting client companies to accelerate AI adoption. Whether in manufacturing, healthcare, or other sectors, the pressure to adopt AI-driven systems has increased. That leads to further dependence on established technology partners. This creates new opportunities. The key question is how well individual companies are positioned to exploit them. Differentiation among companies will increase based on how effectively they enter these newer opportunity areas.
Q: If anything, you are saying there are newer opportunities. But the IT business model rests on two legs — large teams and billable hours. Do you see that getting impacted and affecting revenues going forward?
Pai: First, fixed-price contracts, which are output-based and not based on billable hours, account for 45–50% of revenues for most IT service companies. There could be a push to increase that share. Billable hours are used where there is greater uncertainty. Yes, with AI, tasks can be done faster.
But remember, there is a huge backlog of work in large enterprises. For example, major banks may have 12–14 different databases instead of a single unified database. You cannot effectively run AI in such an environment. They spend $15–17 billion annually on technology and services. That is not going to disappear overnight.
There will be productivity improvements and more work coming in. India is ideally placed because no other country has comparable scale in project management, skilled manpower, and the ability to handle large transitions quickly. Individual project sizes may shrink, but there will be more projects because work can be completed faster and better. Unproductive effort will reduce, and employees will become more efficient.
This is a positive development. When SaaS companies emerged, many claimed they would destroy application development. That did not happen. There are waves of disruption, and AI is a significant one. AI will reduce effort and increase productivity. Costs may come down, but much of the savings will likely be reinvested in improving technology capabilities, because enterprises must modernise and remain competitive.
With $20–25 trillion of accumulated software across global enterprises, there will be substantial transition work. Most global enterprises do not have the in-house capability to manage this transformation. Where is the skill to use and test AI-generated code effectively? Service companies will continue to play a critical role. The situation is overblown. AI will drive change, but Indian service companies will continue to do well and get more work.
Watch accompanying video for entire discussion.
