Where TCS, HCLTech, Infosys, and the rest still win

A long long time ago, I was employed at one of those giant IT-service companies in Chennai, just after I graduated from engineering college.

Astute readers of Zero Shot will note that I wrote, “employed” and not “worked”, because it would be a stretch to describe what I did as work. In fact, my stint at the company was so timepass that I don’t even mention it in my Linkedin history. You wouldn’t blame me either because my IT-services career looked something like this:

Anyway, I was hardly an ideal employee. I skipped most of my sessions to watch terrible movies at Satyam Theatre, which was helpfully located just behind my training centre. So for the next few months, while I was still technically employed, I worked as a reporter at a business daily—because I simply couldn’t stand the intricacies of health-record databases. These were the good old days, when people still fled into journalism. On the plus side, I remain something of a master at COBOL, a language smart monkeys can pick up.

The reason I’ve been thinking about my brief career in IT services is the recent flood of news about how AI is going to kill and destroy IT-service companies. The general narrative is that smaller, AI-native service companies will be much nimbler and move fast to disrupt the incumbent and traditional service companies. There’s already some signs that this is happening. Just last week, Opendoor, a US-based homebuying platform, got rid of its India team and moved it to an AI-native team in the US. Then there are stories of how GCCs (Global Capability Centres) are scaling their AI talent. In fact, I’ve even argued in earlier editions of Zero Shot that legacy IT-service companies are learning the wrong lessons from the past.

Well, I’ve changed my mind about some of this.

The more I think about it, the more I’m convinced that reports of IT-service companies like TCS, Infosys, and HCLTech being outmatched are exaggerated.

In fact, if things continue this way, my prediction is that the service-native AI companies—which is what these incumbents will become—will out-compete the AI-native service companies. It’s not a slam dunk, and there are some uncertainties, but it’s no longer unviable. In fact, if anything, I think IT-service companies have started a process of transformation that will make them stronger and more durable.

My reading comes from connecting a few dots.

Here’s how one company describes what it does. Read it and guess which company it is.

Consider a multi-site healthcare services group, like a network of physician practices. Clinicians spend hours each day on documentation, medical coding, prior authorizations, and compliance reviews. An engagement might begin with the company’s engineering team sitting down with clinicians and IT staff to build tools that fit into the workflows that staff already use. The clinicians know where time disappears in a shift and what good patient care actually requires. The company’s engineers build around that knowledge, allowing clinicians to devote more time to patient care.

Engagements like this will run across mid-sized companies across industries, each shaped by the people closest to the work.

If you guessed TCS, Infosys, HCLTech, or Wipro, well, you’d be wrong.

The idea that the world’s mightiest company is spinning up a service company may seem like a warning bell for the likes of Indian IT services, but if anything, it’s the strongest repudiation that the work they do cannot be easily replicated by an AI-native company, i.e., the activity itself is the moat.

Ask anyone in IT services who their clients are, and there’s a pretty good chance you’ll hear one of three categories: banks, insurers, and financial firms. In fact, BFSI is the single largest vertical at every major Indian IT company. It comprises about 31% of revenue at TCS, 28% at Infosys, and 34% at Wipro. Healthcare (also known as life sciences) is the other name you’ll hear, though it’s a smaller slice (around 14% at Wipro, and about the same at Infosys).

If you’d asked me the same question 20 years ago when I was employed in IT services (assuming you could have torn me away from my gripping COBOL coding sessions), I’d have given exactly the same answer. Some things haven’t changed much.

I used to think that this was a bug, but I’m increasingly coming around to the belief that this is a feature.

Because guess what else is common to banking, finance, insurance, and healthcare?

They are all heavily regulated.

I don’t think we truly understand the enormity of how hard it is to offer software services, literally touching every part of the business, including sensitive stuff like banking and customer records, of some of the most regulated and sensitive businesses in the world. The fact that Indian IT services managed to do this, and continued to do it for decades is pretty incredible. We take it for granted, but it really shouldn’t be. As an example, go and find out how many SaaS or data-provider companies have managed to make deep inroads into such companies, and you’ll find that it’s a pretty short list. A notable exception is Salesforce, and they don’t touch business operations and legacy databases.

A lot of people look at the work that IT-service companies do for their clients and scoff at how AI can do it better. But the real moat was never the work or the service—it was the fact that IT-service companies crossed multiple barriers of compliance, audits, checks and balances, and legal indemnity. Also, all of these legacy companies have a solid balance sheet that lends significant support during selection, because it signals that they can absorb indemnity in case something goes wrong.

A 50-member AI-native startup can move faster, but to do that, it must first get in the door. And that’s really the hardest part of this entire process.

Of course, you could argue that regulatory compliance is a matter of time, and that eventually AI-native service companies would cross those hurdles too. Perhaps.

But from the perspective of the AI companies themselves, they’re also discovering that maybe it’s not worth it.

Anthropic has partnered with Indian IT services giant Tata Consultancy Services (TCS) in a bid to accelerate adoption of its artificial intelligence models at enterprises.

The partnership will see TCS creating a business unit focused on deploying Anthropic’s AI models to its customers. TCS will also gain early access to new model releases, which it says it will use to build expertise, and it will provide Anthropic’s Claude AI assistant to its employee base of more than 50,000 people.

The companies said they would develop solutions for sectors like financial services, healthcare, telecommunications, and aviation.

Anthropic taps TCS to scale its enterprise AI deployments, Techcrunch

Financial services. Healthcare. Telecommunications. Aviation.

What’s common to all of them?

The other common argument I hear from AI-native service companies has to do with business models. The most basic form of this is that IT-service companies bill their customers through time and materials (essentially hourly rates) while AI-native companies will bill clients based on outcomes. This suggests that native companies have the incentive to use AI and agents while IT-service companies don’t. Taken to its logical conclusion, this means that IT-service companies will get disrupted by their business models, while AI-native service companies will win.

This is already happening. Across industries, clients are telling their providers the same thing: use AI, get it done, lower the rates.

You’d assume that the margins of IT-service companies would be under severe pressure and they’d be dropping billing rates to not lose their clients.

Well, then how do you explain this?

Tata Consultancy Services Ltd. unveiled its financial results for the fourth quarter and full year ending March 31, 2026, on April 9, 2026, showcasing a strong quarterly comeback that drove its stock price up 1.16% to ₹2,559.2. The presentation revealed operating margins reaching a four-year peak of 25.3%, even as the IT services giant navigated currency headwinds and macroeconomic uncertainties.

TCS Q4 FY26 slides: margins hit 4-year high amid revenue rebound, Investing

Imagine that. India’s largest IT-service provider didn’t see a reduction in margins, it saw the biggest margin increase in four years. By the way, this story is from April, before the full force of the exchange rate benefit accrued to them.

Margin increases happen because companies are able to sell additional services, likely AI integrations into their deep workflows that they understand intimately, to highly regulated, massive Fortune 500 companies. Since these companies are already embedded, IT-service providers’ ability to increase and charge for integrations and transformations is a cash cow that’s waiting to be exploited, likely for years.

But to do that, they need to rationalise their costs, which will happen through slower hiring and the right mix of AI and humans (which is what TCS has suggested will happen).

Finally, my last argument is that AI unlocks something incredible and special for IT-service companies which they’d never touched earlier.

And it’s all to do with my one area of expertise.

If you don’t know anything about COBOL, let me say that it’s probably the oldest language out there, and used extensively as the backbone of legacy systems like banks, insurers, airlines, and governments in a complicated roundabout way to stitch multiple systems together. It’s old, a pain to use, to debug, and few people know it well. And so IT-service companies hire fresh engineers like me and train them to work on it (while we aren’t skipping to go to Satyam Theatre).

In fact, by one estimate, there are about 800 billion lines of COBOL still in production systems today.

Everybody hates COBOL, but the cost of rewriting and upgrading out of it to a modern language is so prohibitive that clients do not bother to do it. And IT-service companies also cannot offer a way to make the transition without throwing tons of COBOL programmers at the problem, who need to know both COBOL and another language. It’s simply not possible to make the cost-benefit equation work for clients.

COBOL modernisation differs fundamentally from typical legacy code refactoring. You aren’t just updating familiar code to use better patterns, you’re reverse engineering business logic from systems built when Nixon was president. You’re untangling dependencies that evolved over decades, and translating institutional knowledge that now exists only in the code itself.

Modernising a COBOL system once required armies of consultants spending years mapping workflows. This resulted in large timelines and high costs that few were willing to take on.

Tools like Claude Code can automate the exploration and analysis phases that consume most of the effort in COBOL modernisation.

How AI helps break the cost barrier to COBOL modernisation, Anthropic

Essentially, if AI cuts the cost of modernisation, then thousands of legacy projects across healthcare, insurance, banking, and airline companies that were too expensive to ever attempt now clear the bar.

And guess who’ll benefit the most from this?

Will it be an AI-native service company on the outside with no domain knowledge, expertise, or audited compliance with these clients?

Or the one that’s already inside the walls?

All reports of the death of IT-service companies are vastly exaggerated.

Sure, maybe they’re late to the game, and they seemed to be lost. But they are making moves now and some of them are pretty unconventional, as my colleague Rohin reported last week. Of course, many things may go wrong. Things can slow down or companies can face unknown problems. But the pessimism in Indian IT services has less merit than the conventional wisdom.

Expect to be surprised by the likes of Infosys, TCS, HCLTech for the next few quarters.

And if you still don’t believe me, try learning COBOL.

Hi everyone! Vidhatri here. Hope you are getting to rest and recharge this Saturday!

I was listening to my colleague Sakshi Sadashiv tell a story the other day. She was at her friend’s place. The friend’s roommate had come back from a long day at a Big Four consulting firm. He didn’t decompress. Or chill. Instead, as she told us, he strapped a phone to his forehead and started folding clothes. He did it again and again, recording every movement.

That footage would eventually be sold to a data-collection company. Which would sell it to Tesla. Or Figure AI. Or Agility Robotics. All to teach a “humanoid” how to do what he just did.

Sakshi has documented this mushrooming gig economy around data for physical AI in her latest story. I’d urge you to read it. It is sharply-reported and superbly written.

She joined us on the latest episode of Zero Shot along with Pramod Ghadge, the co-founder of Unbox Robotics. Both took us through the different layers of the physical AI ecosystem.

Our host Praveen Gopal Krishnan or PGK entered the chat with his own framing that presented physical AI as a Mexican standoff between three competing players. China has factories. The US has the brains. India has the data.

Who actually ends up with an advantage? Tune in!

You can listen to the episode on Apple podcasts, Spotify, or The Ken app.

Written by Praveen Gopal Krishnan