EXL Q1 2026: AI Revenue Hits 60% and Reshapes the Insurance Vendor Model

EXL reported $570.4 million of Q1 2026 revenue on April 29, with data and AI-led services at 60% of the total, up 28% year over year. The insurance segment grew 12.6% to $193.9 million and management raised full-year guidance. The 60% is the number worth examining, because it measures how engagements are classified and priced rather than how much of the work a machine now does.

Key Takeaways

• 60% of revenue is classified as data and AI-led, roughly $342 million in the quarter against about $268 million a year earlier, an internal classification rather than a GAAP line item.
• Digital operations revenue fell 2% while AI-led grew 28% and total operations still grew 10%, so the shift is additive rather than purely a relabeling of the same book.
• $193.9 million of insurance revenue, up 12.6%, at a 37.7% gross margin against 36.6% a year earlier.
• Pricing is moving to fixed-fee, milestone, and outcome-based structures, which means AI productivity gains accrue to the vendor rather than returning to the carrier as a lower unit rate.
• Revenue per employee reached roughly $34,000 annualized, up from about $30,000 two years earlier, on headcount of around 67,000.

What the Quarter Reported

Revenue of $570.4 million grew 13.8% on a reported basis and 13.4% in constant currency, with 5.1% sequential growth. Adjusted operating margin expanded 40 basis points to 20.5%, adjusted diluted EPS rose 20.2% to $0.58 from $0.48, and adjusted EBITDA reached $127.9 million at a 22.4% margin. The company added 16 new clients and carries roughly 67,000 employees.

Segment	Q1 2026 Revenue	Q1 2025 Revenue	YoY Growth	Gross Margin
Insurance	$193.9M	$172.0M	12.6%	37.7%
Healthcare & Life Sciences	$151.9M	$125.6M	21.0%	45.3%
Banking, Capital Markets & Diversified	$127.4M	$117.7M	8.1%	36.9%
International Growth Markets	$97.1M	$85.7M	10.9%	34.1%
Total	$570.4M	$501.0M	13.8%	38.9%

Guidance moved up to $2.30 billion to $2.33 billion from $2.275 billion to $2.315 billion, implying 10% to 12% growth, with adjusted EPS of $2.18 to $2.23 against a prior $2.14 to $2.19.

The 60% figure needs parsing. EXL does not break out AI-led revenue as a GAAP line. It classifies whole engagements by delivery model: work delivered through AI-augmented workflows counts as data and AI-led, traditional process execution counts as digital operations. At 60% of $570.4 million, that is approximately $342 million against roughly $268 million a year earlier.

The internal composition is the informative part. Reported digital operations revenue declined 2% year over year while AI-led grew 28%, and total operations still grew 10%. The AI category is drawing revenue out of the older category and adding net new demand at the same time, which is what carriers expanding scope into claims triage and underwriting document processing looks like in a vendor's segment data. EXL's Insurance LLM and patent portfolio is the capability underneath it.

The Contract Structure Is What Actually Changed

CEO Rohit Kapoor described the commercial model moving toward fixed-fee, milestone-based, and outcome-based payment. That single sentence carries more consequence than the revenue mix.

Under per-FTE or per-transaction pricing, a vendor that processes more claims per employee hands the saving back at the next contract cycle, because the unit is labor. Under a fixed fee, total cost is locked regardless of how the vendor delivers, so productivity improvement lands in the vendor's margin. Outcome pricing goes further and ties fees to cycle time or accuracy targets.

The margin data is consistent with that reading. Adjusted operating margin expanded to 20.5% while revenue grew only about three percentage points faster than headcount. Under a pure FTE model that spread would be competed away at renewal.

For the carrier, the change lands in the expense ratio rather than the loss ratio. Input-based vendor pricing was forecastable to a high precision because it was headcount times rate. Outcome-based pricing varies with delivered performance, which introduces a source of expense variability into a line that expense analyses have historically treated as fixed and known. It also transfers productivity risk to the vendor, who absorbs the cost of supplementing with human labor when the models underperform.

The peer set is not measuring the same thing. Verisk reported $783 million, up 4%, with subscriptions at 84% of revenue and seven new AI modules against a 25-module full-year target; it discloses no AI-specific revenue because its whole base is analytics. Guidewire grew 22% to $356.6 million on cloud migration, treating AI as a feature inside core platform subscriptions.

Sapiens ran roughly $595 million in 2025 with recurring revenue at 79% and a 2% to 3% headwind from its SaaS conversion. EXL is the only one publishing an AI revenue mix, which is a disclosure choice as much as a business difference.

The Reclassification Is Running Ahead of the Automation

The productivity evidence is thinner than 60% suggests. Revenue grew 13.8% against headcount growth of approximately 11%, a three-point spread. Annualized revenue per employee works out to roughly $34,000, up from about $30,000 two years earlier, at a company that still employs around 67,000 people.

That is genuine improvement and it is not the picture a majority-AI revenue share implies. EXL's own description of the model is human-in-the-loop AI-led, and the classification counts an entire engagement as AI-led when AI is the primary delivery mechanism even where human operators remain in the workflow. The 60% therefore measures a commercial and contractual migration, with the delivery migration trailing behind it.

That gap matters most where the outputs land. A carrier using AI-augmented claims processing takes the results into its reserving data, and the classification tells the carrier nothing about which parts of the workflow a model touched. If a triage model drifts, the effect enters the loss triangle through a vendor boundary the carrier cannot inspect, and the contract that produced it is priced on outcomes rather than on the steps that generated them.

The switching cost is deliberate. Moving from an outcome-based engagement to an internal capability means rebuilding the performance measurement framework alongside the technology, which is a harder exit than replacing a per-transaction supplier. Celent's third annual survey puts 48% of insurers in production with generative AI and 22% expecting an agentic solution by the end of 2026, most of them arriving through vendors rather than internal builds, and the NAIC evaluation tool pilot will be asking those carriers to document capabilities that sit inside somebody else's classification scheme.