EXL Grew 20%, Raised Guidance, and Admitted It Can't Yet Price the AI It Sells

A 56% ninety-day rally into the print, a fourth straight insurance acceleration — and a CEO openly conceding the commercial contract hasn't caught up with the productivity he's delivering.

EXLS · Earnings Call · 2026-10-28

The quarter the reframing finally paid

ExlService Holdings printed $634 million of revenue, up 20% reported and 17% organic, with adjusted EPS of $0.62 up 29%. Every one of the four segments grew more than 15%. Guidance went up — again — to $2.43–2.44 billion and $2.34–2.36 in EPS, the second raise of the year. Rohit Kapoor was blunt about the shape of it: “Organic growth accelerated to 17%, with every segment growing more than 15%, reflecting broad-based momentum in our business that has strengthened throughout the year.” — Rohit Kapoor, Chairman and Chief Executive Officer · 2026-10-28 That is a genuinely good quarter for a $6.2 billion IT services name. But the interesting part is how the numbers are being assembled, because the revenue line has been quietly re-plumbed. The headline growth is coming from the data and AI bucket — up 36% reported, 31% organic, now roughly two-thirds of revenue. The legacy bucket, Digital Operations, shrank 1.7%. Management's preferred container for the whole messy middle is Total Operations, which grew 12%. That's the trick: bundle the declining piece with the booming piece and report the bundle. Fair enough — it's economically honest — but it means the reported 20% is a blend of a mid-teens grower and a shrinking one, not a uniformly accelerating business. The keyword trail confirms the migration. In the prior quarter's curated set, AI led operations and "AI capability" were load-bearing concepts; this quarter "AI capability" is the single largest momentum decliner in the company's own keyword history, and "AI led operations" has fallen off hard too. What replaced it is vocabulary of contracts and money: commercial construct, business value, organic basis. The story has moved from we can build it to we can't yet bill it the way we'd like.

The admission that matters

Maggie Nolan of William Blair asked the only question that really counted: what does the deliberate migration of clients from traditional operations into AI-led work do to the margin profile of those engagements? The answer, unvarnished:

The margin story on this is a little bit unclear at the current moment because the change in the commercial construct with our clients, that is really not shifting as rapidly as the embedding of AI and data into the operations. Our clients still prefer us to kind of stay with existing commercial constructs.
Rohit Kapoor, Chairman and Chief Executive Officer · 2026-10-28

Read that against what the same CEO said six months earlier, when the commercial model was described as already pivoting: “the commercial model is changing much more towards a fixed-fee and milestone-based payment and an outcome-based model.” — Rohit Kapoor, Chairman and Chief Executive Officer · 2026-04-29 It clearly hasn't, or at least not at the speed the equity was told to expect. The global tape already senses the risk in that framing — the theme outcome based pricing sits among the worst 360-day decliners in the market-wide keyword set, and Booz Allen's version of the same idea was flagged in the last five days of reporting. Meanwhile the arithmetic of AI productivity is deflationary, and EXL has been saying so for a year. A year ago: “35% of the transactions, which are AI-enabled, and that's generated 30% productivity benefit for this client. And yet our revenue for this client has remained the same.” — Rohit Kapoor, Chairman and Chief Executive Officer · 2025-10-29 When you hand clients efficiency and hold revenue flat by winning more work, your revenue is a function of how fast you can absorb displaced scope. That is a treadmill, and treadmills compress margins before they expand them. Hence adjusted operating margin of 19.3%, down 10 basis points, and a guide for roughly flat margins. The GAAP trail is equally unrewarding: operating margin of 14.7% at the last reported quarter, down 1.1 points year over year.

What management stopped talking about

Vivek Jetley is gone. The President who ran Insurance and Health & Life Sciences left in September, and the tell is that he was the voice explaining the Payment integrity thesis on both the February 2026 and July 2025 calls — the payer cost-pressure tailwind, the Q4 win with a top-five US payer, the argument that payers wouldn't take the work in-house. This quarter Kapoor answered it himself, and the answer was better than the succession optics: “instead of us looking at a certain percentage of the claims that a healthcare payer might be making, we're taking a look at 100% of the claims that go through the system.” — Rohit Kapoor, Chairman and Chief Executive Officer · 2026-10-28 Extending payment integrity from healthcare into insurance claims is a genuinely large TAM claim, and unlike most TAM claims it comes with an entirely outcome-based revenue model already in place. Still, the keyword decay is worth flagging: "Insurance segment" momentum fell sharply this quarter, and the insurance narrative has been handed to Bhupender Singh, whose International Growth Markets patch grew 21% — accelerating from ~13% in the first half. That segment is the smallest of the four and the most underpenetrated, and Kapoor has now twice hinted at acquisitions there.

iMerit, frontier labs, and the tape's verdict

iMerit contributed $15.3 million in two months of ownership. Kapoor was refreshingly unwilling to spike the ball: “The integration has gone off really well, but it's still too early to call this a success.” — Rohit Kapoor, Chairman and Chief Executive Officer · 2026-10-28 The strategic logic — playing across the full AI stack including model training, evaluation and agentic assurance — shows up as a consensus theme in this week's reporting cluster: Corning's CLS flagged a frontier AI labs customer, Comfort Systems named a frontier lab, and Infosys leaned on AI-led program wins. When three unrelated reporters in five days use the same phrase, that's a sector signal, not a company one. The market, though, is not paying for sector signals right now. Global price action shows AI data center up 53% over 360 days but down about 5% over both 90 and 30 days — the infrastructure trade has gone stale. EXL, an application-layer play on the same thesis, is the mirror image: +55.7% in 90 days, closing at $42.91 as recently as October 22, only 5.1% off its high, while still 21% below the February 2025 peak of $51.80. The re-rating is real but incomplete — price to net income of 15.7x, down 48% year over year, and price to revenue at 1.8x off a 2022 peak near 20x. What funds the next leg matters too. effective net cash has swung from +$45M at year-end 2025 to −$458M at mid-year 2026; management now carries $513M of term loan and revolver debt against $333M of cash, with borrowing capacity expanded to $1 billion and liabilities to assets at 50.4%, up 12.5 points year over year. They bought back $205 million of stock in nine months. A company simultaneously buying stock, buying capability, and telling investors margins stay flat is asking for patience on two fronts at once.

The bottom line

EXL has the best top-line story in IT services this quarter and the vaguest margin story. The Agentic AI adoption claim is now backed by numbers, not slides — 90%+ engagement success rates, 30 new logos, double-digit pipeline creation, and a 15% organic constant-currency guide for the year. But the spread between headcount growth and revenue growth has widened to roughly 600 basis points from the ~3% "steady state" management described a year ago, and that gap is the entire margin question in one number: either it becomes operating leverage as the contracts reprice, or it becomes the price of staying relevant. Kapoor put the timeline at a couple of years. The stock has already put the timeline at a couple of quarters.