EXL Is 60% an AI Company. Its Stock Is Still Priced for Disruption

The timing is almost too neat to be coincidence. On the day EXL marked two decades as a public company, it rolled out a new brand, "Go Beyond," built around a claim that would have sounded absurd at its IPO two decades ago: more than 60% of its revenue is now data- and AI-led. That is a statement about a twenty-seven-year-old firm's core, not its pitch deck. And yet the market keeps treating EXL as part of the old world. The shares trade around $35, down for the year and roughly a quarter below their $47 high, even after a 17.9% one-day jump when second-quarter results beat expectations and management raised its own guidance.

That gap — an improving business, a punished price — is the real story, and it is the whole debate in miniature.

The reinvention was real, not cosmetic

EXL began life in 1999 as a data-intensive business-process operator for insurers and banks: analytics first, then outsourcing. For most of its history that was a labor-arbitrage model — clever people in lower-cost geographies doing work clients used to do themselves. The "Go Beyond" launch is management's way of saying the mix has flipped. The company no longer describes itself as an outsourcing firm; it describes itself as a global data and AI company, and the 60% figure is the load-bearing number, the point at which a business stops being "an outsourcing firm with some AI products" and becomes "an AI business with a services backbone."

The recent numbers support the repositioning, and they matter less for the quarter than for what they say about direction. Second-quarter revenue came in at $594.8 million, up 15.6% from a year earlier, with adjusted earnings per share up 22.3% to $0.59. That is not a company standing still while the industry frets. Management has raised its full-year outlook twice in a matter of months — from 10% to 12% growth laid out at Investor Day in May to 14% to 16% now, with adjusted EPS guided 16% to 18% higher.

What is underneath that growth is just as telling. EXL has become an ecosystem plug: an OpenAI services partner, part of Anthropic's Claude partner network, a Databricks and Snowflake partner, and integrated with NVIDIA's transaction foundation model. And in August it closed the acquisition of iMerit, a specialist in training and evaluating AI models — not legacy outsourcing work, but the grunt labor of making frontier models reliable. It is a bet that the value in AI accrues to the people who make it work inside messy real-world businesses, not to the model labs.

Why the market is still squeezing the sector

The pessimism is not aimed at EXL in isolation. It is aimed at the whole advisory-and-service layer. When Gartner — the industry bellwether — reported consulting revenue down 12.8%, the entire complex sold off on a specific fear: that generative AI compresses the billable-hour model, that clients will need fewer consultants and operators because the models do the drafting and the analysis.

For EXL that fear is existential rather than academic, because its economic model is precisely the one being called into question. A skeptic looks at "60% AI-led revenue" and sees the same engagements relabeled — the analytics still delivered by people, the AI lipstick on the outsourcing lipstick. That is the honest worry, and it is why a services name trades at a fraction of the multiples the AI-native names command.

The challenge is that the two readings produce the same 60% figure. Data/Q2 beats and raised guidance tell you the mix shift is producing results today. They do not tell you whether those results survive once models get good enough to do more of the work autonomously. EXL's own research frames its opportunity in exactly this gap: a study it published found 76% of companies believe they are ahead of their competitors on AI, yet only 10% qualify as genuine AI leaders. The distance between what executives believe and what their operations actually deliver is the business EXL is selling.

Where the divergence leaves you

Here is the contrarian machinery worth paying attention to, because it is a genuine divergence and not a claim I'm forcing: the lead indicators — the numbers management controls and raises — have been pointing one way for two straight quarters, while sentiment has been pointing the other way all year. After the Q2 beat, after the raised guidance, the stock still trades at a trailing multiple around 22 and a forward multiple near 14, with a PEG under one for a company growing revenue 14% to 16%. That is the shape of a market pricing in disruption that has not yet shown up in the company's own results.

A setup this divergent usually resolves toward whichever direction the data actually owns, and right now the data belongs to the raised guidance. But conviction here is not automatic. The bull read is invalidated if the growth engine stalls — if guidance rolls over, if the iMerit integration underdelivers, or if the "AI-led" revenue growth decelerates below the corporate average and reveals itself as branding. A rebrand at twenty years public is cheap marketing. The strategic bet underneath it is expensive to execute: turning a labor business into an intelligence business fast enough that the curve doesn't eat the old model first.

That is the real question the anniversary party obscures. AI adoption is on an exponential curve; the only question is whether EXL is on the right side of it. The company's own numbers say it is — for now. The rest is whether the market's fear or the company's guidance is the better forecast, and the evidence for the fear is currently thinner than the evidence for the guidance.