Kapoor’s Curiosity for Context Keeps EXL Focused on Data
Analytics India Magazine (Mohit Pandey)
When it comes to AI, the world is obsessed with model makers like OpenAI and Anthropic, or the big-tech infrastructure players like Google and Microsoft. But, the real winners in this AI boom emerge at the end of each financial quarter, when earnings reveal how a company used AI for real-world problems.
These are the IT, services, and consulting firms. One of them, a smaller player, is EXL, which outperformed even the larger firms this quarter in terms of revenue growth.
The company’s latest quarter set the stage by posting $529.6 million in Q3 revenue, up 12.2% year on year, powered almost entirely by the momentum of its data and AI business. Chairman and CEO Rohit Kapoor said in the earnings call that their data and AI-led revenue grew 18% year-over-year to mark the third consecutive quarter of accelerated growth.
EXL won 21 new clients, rolled out EXLdata.ai with Databricks, expanded its AI-driven CX work with Genesys and pushed its agentic platform EXLerate.ai deeper into the enterprise.
“As the world moves towards greater adoption of AI, how can we help our clients leverage AI?” Kapoor asked AIM rhetorically. He said EXL knows how to put AI inside the workflow, and that is where the real shift is happening.
The EXL Strategy
Kapoor kept returning to the phrase “AI in the workflow.” For him, this is not a slogan. It is the hinge on which EXL’s entire strategy rests. He broke it down with the clarity of someone who has explained this a thousand times. Apart from the data and the model, the context of how a business actually works is important in his view.
“That contextual knowledge of the business and the process and the operation is what allows us to be a superior implementer of AI in the workflow,” he said.
Kapoor believes his curiosity for “context” comes from having spent his career in the guts of insurance, healthcare, banking—the messy sectors—with broken forms, strange edge cases, outdated systems, and human judgment layered over years of exceptions.
EXL’s business, in his view, splits into two clean halves. One where they run operations for clients, and the other where they manage data and build analytics and AI systems for clients. The second half is now 56% of the revenue. “The demand side of that is enormous,” Kapoor said, adding that he expects it to grow in high double digits for years.
According to Kapoor, EXL’s biggest bet is on inconsistent and scattered data across platforms. “We have created EXLdata.ai, an agentic platform that enables this kind of data modernisation in a much quicker and less expensive way.”
He claimed EXL’s work in this domain is 50% cheaper, and delivered in weeks, instead of months.
Even with that, enterprise AI is still mostly stuck in pilots. Kapoor tried to unravel why? Every vertical process is a chain of disconnected systems that don’t talk to each other. Every AI deployment becomes a point solution glued on top of an old system. One step can be automated, but not the entire journey until everything connects.
And in most companies, nothing connects cleanly. The only AI use cases scaling across the enterprise are horizontal ones which include customer support, agent assist, and coding. Things that sit above the business unit and plug into everybody. “These horizontal things have got a lot of scale,” he said.
And he is right. These are the only places where AI is already everywhere.
The Agentic AI Question
Even though everyone is talking about agentic AI, the ROI is still not there, and everyone knows that. The problem Kapoor says is that the first pass of an agent solves tasks at 50-60% accuracy, which is “useless.” It creates more work than it reduces. The real skill is moving it to 98%.
That is where process knowledge matters more than anything. And this is where he believes IT services firms fall short. “They understand the technology part really well, but they don’t know how to iterate on this,” he said, while speaking about the viral MIT study on how 95% of the AI deployments fail during production.
He said EXL’s success rate is 93-95%. They get paid only when the business outcome is delivered, so the incentives are clean. If they fail, they don’t earn. Their margins reflect that.
And to move away from these questions of margin and focus on massive and flashy AI announcements, Indian IT giants like TCS, Infosys, Wipro and HCLTech are announcing hundreds of AI agents and data centres.
TCS, for example, is investing $6 billion on an AI data centre.
But Kapoor shook it off saying that is not EXL’s game—no data centres, no capex. “We are going more and more on the data side,” he said, while explaining that If AI has layers, EXL wants to own the data layer and the fine tuning layer, not the GPU layer.
The company has already built nine small language models tuned for insurance, healthcare and claims. He believes the real moat sits there. He also dismissed the idea of building products like ChatGPT without hesitation. EXL wants solutions, not products. Something 80% ready with 20% customisation.
“Just take what you have right now and make it useful” is what Kapoor’s and EXL’s mantra is in this AI race. About India’s AI talent, he said India must raise its math game. Pure math will decide who builds the next generation of AI, not coding, he claimed.
Kapoor said that AI tools like Lovable and Cursor are pragmatic, helpful and fun, but developers still review every line. The real breakthrough will come when code blocks become verifiable and trusted. At EXL, he has built his own agents inside their sandbox without knowing how to code.
AI is moving fast. Enterprises are moving slowly. The risk is not that AI fails, but that firms misapply it at scale. The companies that win will be the ones that understand the workflow, fix their data estate, and move from 60% to 98% before everybody else.
Kapoor believes EXL is one of those companies.
The post Kapoor’s Curiosity for Context Keeps EXL Focused on Data appeared first on Analytics India Magazine.
Generated by RSStT. The copyright belongs to the original author.