Verisk beat on both lines in Q1 2026: revenue of $782.6 million, up 3.9%, adjusted EPS of $1.82 against a $1.74 consensus, and subscription revenues up 7% on an organic constant currency basis.

The disclosure that had not appeared in prior quarters came from CEO Lee Shavel, who described "additional complexity in negotiating and adapting our contracts" over intellectual property, privacy and AI governance provisions, extending the sales cycle on Verisk's most advanced products. He was explicit that it is not a demand problem.

Key Takeaways

  • Subscription revenue is 84% of the total and grew 7% on an OCC basis, while the 16% transactional base fell 6.1% on low weather activity against a comparison quarter carrying Helene and Milton claims.
  • The friction sits in legal and compliance review, not technical evaluation, which means the constraint on insurer AI deployment is contracting infrastructure rather than whether the models work.
  • AM Best found only 41% of more than 150 rated insurers and MGAs actively using AI across core business areas, against nearly 60% expecting AI to transform their business within one to three years.
  • Fewer than 20% report reaching an advanced implementation stage, with data readiness, security and privacy concerns at 43% and legacy system integration at 41% named as the top obstacles.
  • The products clearing fastest are the ones that do not touch a rating or coverage decision, which is precisely where the friction concentrates the delay.

The Print, and the Sentence Inside It

The quarter reads clean, and the disclosure that matters is not in the numbers.

Metric Q1 2026 Q1 2025 Change
Total Revenue $782.6M $753.0M +3.9%
OCC Revenue Growth 4.7% – –
Subscription Revenue Growth (OCC) 7.0% – –
Transactional Revenue Growth (OCC) −6.1% – –
Adjusted EPS $1.82 $1.73 +5.2%
Adjusted EBITDA $436.0M – +5.9% OCC
Adjusted EBITDA Margin 55.9% 55.3% +60 bps

Revenue of $782.6 million rose 3.9% from $753.0 million, 4.7% on an organic constant currency basis, with adjusted EPS of $1.82 beating consensus by 4.6% and net income up 1% to $234 million. Adjusted EBITDA grew 5.9% OCC at a margin of 55.9%, up 60 basis points, and full-year guidance was reaffirmed at $3.19 billion to $3.24 billion of revenue and $7.45 to $7.75 of adjusted EPS.

The mix is the durable part. Subscription revenues are now 84% of total and grew 7% OCC. Transactional revenue, the remaining 16%, fell 6.1% OCC on low weather activity against a prior-year quarter that carried Hurricanes Helene and Milton claims volume, which is a comparison effect rather than a competitive loss.

Underwriting Solutions grew 5.3% OCC and Claims Solutions 3.4%, with the underwriting side carrying seven new client-facing modules shipped under Core Lines Reimagine against a 25-module target for the year. That program digitizes the forms, rules and loss costs franchise into machine-readable assets, which is the difference between an actuary pulling ISO loss costs and classification codes through an API and looking them up by hand.

Five Review Functions That Did Not Exist Two Years Ago

The friction Shavel described is worth taking literally, because where it sits in the sales cycle determines what it predicts.

Technical evaluation, whether the model works and improves a loss ratio, is often the fastest stage of an AI procurement. What follows is a multi-party review. Legal negotiates intellectual property ownership of model outputs, data rights, indemnification for algorithmic error and liability allocation for AI-generated decisions. Compliance tests alignment with the Colorado AI Act, the NAIC model bulletin and emerging model law provisions, particularly unfair discrimination testing and documentation. Privacy covers data usage rights, cross-carrier sharing restrictions and consent frameworks. Model risk governance covers explainability and audit trail provisions. Procurement renegotiates terms that were boilerplate for a data subscription.

Each runs on its own timeline and approval chain, so the delay is the cascade of five or more independent functions reviewing provisions that did not exist two years ago. Shavel expects it to improve as industry standards for AI governance contracting mature, which is a statement that the bottleneck is external to Verisk.

The readiness data says how long that will take. AM Best's early-2026 segment report, surveying more than 150 rated insurers and managing general agents, found nearly 60% expecting AI to significantly transform their business models within one to three years while only 41% are actively using AI across core business areas and fewer than 20% report an advanced implementation stage. Two thirds plan to increase investment over the next 12 to 24 months.

The named obstacles are the ones that lengthen a contract review rather than a build. Data readiness and quality lead, followed by security and privacy at 43% and legacy system integration at 41%, with legacy systems storing data in inconsistent formats lacking standardization. A carrier that cannot document exactly what data feeds a vendor's model, how it is governed and what consent framework applies cannot close the compliance review, whatever the model does. Verisk's structured-data position, built over decades of industry collection and standardization, helps at its end of the contract and does nothing to the carrier's internal review clock.

The Friction Falls Where the Value Is

The complication is that the products moving fastest and the products carrying the most actuarial weight are different products.

Digital Media Forensics is the counter-example that isolates the cause. It onboarded its sixth top-10 carrier in Q1, using AI to detect manipulated images, deepfakes, reused photos and tampered PDFs in claim submissions, and it is not experiencing the same contracting delay. It also does not set a rate, decide coverage, or produce an output a regulator will examine for unfair discrimination. Nothing in the five-function review above binds hard on a tool that flags a doctored photograph.

The augmented underwriting pipeline is where both the demand and the delay sit. Verisk reported over 20 follow-up meetings scheduled on augmented underwriting solutions, so carriers are actively evaluating AI-assisted risk selection and pricing and are held at the contracting stage rather than the evaluation stage. Those are the products that touch the rate.

The co-development win sharpens the same point. Verisk won a competitive RFP as strategic partner for a global insurer building a digitally native underwriting entity, contributing data, actuarial capability and AI platforms to the operating model rather than licensing software into it. That is a carrier concluding it has the domain expertise to shape an underwriting platform and not the engineering capacity to build one.

It also leaves the model boundary unresolved for whoever signs at the end of it. In a licensed product the vendor's component is identifiable and the carrier's is separate. In a co-developed underwriting entity the design decisions, the data, and the model are jointly produced, so the actuary who has to understand the model and its limitations is documenting a system whose provenance is shared. That is a governance question the contract has to answer rather than a technical one, which is the same reason the contracts are taking longer.

Further Reading on actuary.info

Sources