Coding to AI returns: IT's moment of truth as tech reshapes services

Tech firms shift focus from labour arbitrage to outcomes as AI does most of the work

Artificial Intelligence (AI) is omnipresent. Anthropic Claude, Gemini, ChatGPT, Devin, the names that were barely recognisable a few years ago now define industry after industry. Information technology (IT) services — India’s growth engine and the mainstay of the $300 billion IT services industry — has been among the most impacted. Marquee stocks, including Infosys, Wipro, TCS and HCL, have crashed from their historic highs. The Nifty IT index is down 29 per cent so far this year up to June 29. 

The questions are growing louder in boardrooms from Bengaluru to Boston: If Anthropic Claude Code, Cognition Labs Devin, GitHub CoPilot and others can write code, test software, answer queries and build applications, what role is left for an industry that employs around 6 million in India. 

For nearly three decades, Indian IT services thrived on a simple idea: Move work offshore, add engineers, and deliver projects at scale and low cost. Revenue growth was closely tied to headcount growth. More engineers meant more billing hours. And just last fortnight Tata group Chairman N Chandrasekaran, addressing TCS’ 31st annual general meeting, said the next three years will see as many human employees as AI agents. 

The linear manpower and revenue growth model has been dented by AI. On June 23 at Infosys’ 43rd AGM, chairman Nandan Nilekani said: “AI will not replace a company like ours. It will amplify those who move with purpose and adapt with speed.” Infosys is working with 90 per cent of its top 200 clients on AI initiatives. 

Nitin Rakesh, CEO, Mphasis said, “AI-native companies will undoubtedly play a significant role in advancing models and platforms. A model by itself is a powerful tool, but enterprises operate in complex environments. The real challenge is integrating AI into those environments and translating capability into business outcomes.” 

Vic Gupta, executive vice president, Coforge said, “the centre of gravity has moved.” For decades clients bought capacity, seats and effort, driven by labour arbitrage. “Now they buy outcomes with the explicit assumption that AI does most of the work.”

Infosys’ Nilekani estimates the AI-first services market opportunity could reach $300-400 billion by 2030, referring to services designed around AI from the ground up, rather than simply adding AI tools onto existing systems. 

“The IT services model is shifting from scaling effort to scaling intelligence,” said Nitin Bhatt, technology sector leader, EY India. AI is improving productivity across software development, testing, maintenance and IT operations, allowing firms to deliver more with fewer people. Revenue growth, Bhatt said, “will decouple from headcount increase. Up to a third of the core business of technology services providers faces contraction risk from agentic AI.” 

The change is already visible across the delivery stack. Whether it is application development, infrastructure management, product engineering or support services, AI is now embedded in almost every workflow. Developers use AI-assisted coding tools. Product engineers use AI to accelerate design, implementation and documentation. 

Gaurav Vasu, CEO, UnearthInsight, a Bengaluru-based consultancy, said, “The underlying IT services model has not fundamentally changed. What has changed is the way these services are delivered, with integration of AI into every service offering.” Clients continue to buy managed services, application development and infrastructure support through traditional contracts. AI is becoming a default operating layer, woven into application development, testing, cloud engineering, cybersecurity and support services, making the distinction between AI revenue and non-AI revenue increasingly meaningless. 

At Tech Mahindra, around 70 per cent of engagements incorporate AI in some form — whether through software engineering, customer experience, analytics or enterprise operations. Nikhil Malhotra, CIO & global head of AI and emerging technologies Tech Mahindra said, “AI is being embedded across transformation programmes, enabling organisations to improve efficiency, enhance decision making and unlock new growth opportunities.” 

The bigger change, however, is taking place in client conversations. “The conversation with clients has moved from effort to outcomes. Boards are asking for AI return on investment,” said Rakesh of Mphasis. 

Clients no longer want vendors to merely deploy tech — they want partners who understand business problems, take accountability for measurable outcomes and remain invested in delivering results. 

Intelligent engineering 

Traditional time-and-material contracts (where clients pay for labour hours used, and software) are giving way to platform-led offerings, recurring revenue models and outcome-linked engagements. AI is reducing the cost of producing software by 20-50 per cent, according to studies by McKinsey and Gartner. 

As software becomes cheaper, the premium shifts to understanding business context, redesigning processes and integrating AI safely into complex enterprises. 

But much of the anxiety around AI delivering IT services stems from the rapid rise of AI-native companies such as OpenAI, Anthropic, Gemini and their increasingly capable models. Industry executives believe they will get better at executing tasks.

Sure, an AI model can generate code. Yet, it cannot, by itself, modernise a bank’s core systems, integrate with thousands of applications, satisfy regulators or redesign business processes across dozens of countries. “Enterprises don’t run on raw model capability. They run on messy, regulated, deeply contextual systems of record that took decades to build,” said Gupta of Coforge. 

That layer is where Indian IT companies believe they will continue to create value. “AI-natives supply the intelligence. We supply the engineering, domain depth, governance and operational accountability that turn intelligence into a running enterprise outcome,” added Gupta. 

Vasu argues that enterprise software giants such as SAP, Oracle, Salesforce and ServiceNow could automate significant portions of implementation and maintenance work by embedding AI directly into their platforms. AI-powered development tools are simultaneously reducing effort across coding, testing and infrastructure management. 

That will shrink parts of the traditional services pie. The winners will be those who move beyond “AI for IT, towards AI for business,” says Bhatt. 

For Mphasis about 64 per cent of new total contract value (TCV) wins during the latest quarter were AI-led. Overall deal wins jumped 68 per cent year-on-year to $2.1 billion, while the average large-deal size rose sharply. Capgemini reports that generative and agentic AI accounted for more than 11 per cent of its global bookings in the first quarter of calendar year 2026, reflecting what it describes as strong commercial momentum. 

Padmashree Shagrithaya, executive vice president, Capgemini India said, “2026 is the year of truth for AI. The focus is shifting from experimentation to scaled, measurable business value. 

Build, buy or collaborate 

To prepare for that future, IT companies are investing heavily in proprietary AI platforms rather than relying solely on third-party models. Infosys has Topaz, Wipro has ai360, Mphasis has Tria, Coforge has AgentSphere and Forge-X. Acquisitions are also becoming more targeted. Rather than buying AI-native companies outright, firms are acquiring capabilities in data engineering, cloud transformation, cybersecurity, digital engineering and industry-specific platforms that strengthen enterprise AI implementation. 

Services providers like Capgemini are also collaborating across the AI ecosystem. Shagrithaya said, “A key example of this collaboration is our Frontier Alliance with Open AI alongside a strategic investment in OpenAI Deployment Company

(also called DeployCo). We are co-building bespoke industry specific AI solutions and building a dedicated delivery function with OpenAI certified experts.” 

The alliance focuses on embedding AI coworkers directly into corporate workflows. DeployCo, a unit of OpenAI launched in May 2026 is a joint venture between OpenAI and 19 global entities (including Capgemini, McKinsey, and Bain & Company). The goal is to bring full-scale AI-led production capability across clients. 

Over the next few years, AI agents will become routine members of enterprise delivery teams. “The delivery teams will become a small group of humans, directing a large fleet of AI agents,” said Gupta. 

Humans will focus on judgment, context, exception handling and governance. In fact, the era of recruiting thousands of fresh graduates every year will give way to a smaller, specialised workforce, as TCS’ Chandra sees it. The industry’s next chapter will not be measured by how many engineers it employs but by how effectively it combines human expertise with machine intelligence. 

In an AI-driven world, understanding business context, managing risk and knowing where human oversight matters may prove more valuable than millions of lines of code — much of which, AI will do. 

AI models and their capabilities

• Cognition Labs’ Devin: Devin is like a cloud-hosted autonomous software engineer. It can run whole IT projects. Fortune 500 companies are using Devin, among other things, to modernise legacy code
• Anthropic’s Claude Code: Claude Code helps with reasoning-heavy tasks. It is like having a skilled junior programmer living inside your computer. Instead of just suggesting lines of text as you type, it can handle entire programming chores from start to finish, fix bugs, update old files and clean up messy ones
• GitHub’s Copilot: It is a code-completion helper inside your editor. It can complete tasks for you, like translate code, complete sentences, write tests, etc. Instead of doing chores for you like a separate assistant, it sits quietly next to you as you type and instantly guesses what you want to write next

---

The writer is a New Delhi-based independent journalist