Verisk
Verisk is a publicly traded data analytics and technology provider to the global insurance industry, listed on Nasdaq and a constituent of both the S and P 500 and the Nasdaq 100. Through the Insurance Services Office, formed in 1971 and a wholly owned subsidiary since October 2009, it operates as a licensed advisory organization and statistical agent: it collects industry wide policy and claims data, projects future claim costs, and submits standardized forms and loss costs to state regulators on behalf of insurers. Those filings form the starting point from which most United States property and casualty carriers build their own rate filings. Its published data scale is 34.5 billion statistical records, comprising 8.2 billion commercial lines and 21.5 billion personal lines records, and its coverage language is maintained against roughly 10,000 legislative bills, 8,000 regulatory actions and 2,000 court decisions each year.
The business reports in two groupings. Underwriting covers forms, rules and loss cost services plus extreme event and catastrophe solutions used in reinsurance and insurance linked securities. Claims covers anti fraud and casualty solutions alongside the Xactimate estimating ecosystem.
AI is embedded into products carriers already run rather than sold as a separate line. Generative capability is live in the premium audit research tool, the form composition tool, the estimating platform and the claims document tool, with a claimed reduction of up to 98 percent in research time on premium audit documents. XactAI spans claim summaries, photo labeling and description generation, document data extraction, plain language estimate construction, inventory pricing and estimating recommendations, and integrates across the Xactimate, XactAnalysis, XactContents, XactXpert and ClaimXperience products. Licensee numbers rose nearly tenfold between March 2026 and mid 2026 to approximately 7,000. Digital Media Forensics detects image manipulation and deepfakes in claims. Seven Core Lines Reimagine modules shipped in the second quarter of 2026 against a stated target of 25 for the year.
In May 2026 the company launched two Model Context Protocol connectors built with Anthropic for underwriting and claims, bringing its datasets into Claude workflows with retrieved data stated to remain inside the client session and outside model training, alongside a Claude Enterprise integration giving natural language access to loss cost trends and estimating intelligence.
Governance is published rather than asserted. A Commitment to Ethical and Responsible AI disclosure sets out ethical principles, an AI governance structure, standards and review processes, and the claims document product's governance has been aligned to the state commissioners association model bulletin and the NIST AI Risk Management Framework through an independent governance vendor's control library.
Second quarter 2026 revenue was $806.3 million, up 4.3 percent, with subscriptions at 83 percent of revenue and full year guidance of $3.19 to $3.24 billion. Headquartered in Jersey City, New Jersey, with teams across more than 20 countries.
Capability Axes
Capability grades
15 of 15 axes rated · 11 graded A or B
A licensed advisory franchise of 34.5 billion statistical records, standard coverage forms, filed loss costs and the claims estimating platform would stand without any model. AI is placed inside workflows teams already run, which is embedding rather than dependence.
Claims outputs are recommendations, estimates, line items and summaries in front of an adjuster, designed to assist rather than replace. In underwriting the AI does the initial review before a person takes over. That stance lives in product marketing, with no autonomy modes, referral triggers or acceptance thresholds a carrier could configure.
Loss cost projections built by staff actuaries are filed with state regulators, which subjects the modeling to external review. Accuracy for photo labeling, document extraction and estimating recommendations, model cards and drift monitoring for the generative features are unpublished.
Securities filings audit the figures: second quarter 2026 revenue of $806.3 million, up 4.3 percent, with full year guidance of $3.19 to $3.24 billion. Claims AI licensees rose nearly tenfold to about 7,000 by mid 2026, a count of licensees rather than outcomes.
A published Commitment to Ethical and Responsible AI sets out a governance structure and review processes inside corporate risk machinery, beside a dedicated data stewardship function. Data retrieved through the Model Context Protocol connectors stays in the client session and out of model training, a precise and checkable assurance.
Carriers' loss and policy experience is pooled by a licensed statistical agent under periodic state examination, which puts the pooling under supervision rather than private terms. A data processing agreement, retention schedule and subprocessor list are still absent.
Trust center, attestation report and security certification are all absent for a listed company holding industry wide claims and policy data. Securities disclosure rules now reach cybersecurity risk management, yet nothing a buyer can assess is published before a confidentiality agreement.
Its standing is statutory: a licensed advisory organization and statistical agent under state regulation, filing forms and loss costs with regulators on insurers' behalf. Coverage language is maintained against roughly 10,000 bills, 8,000 regulatory actions and 2,000 court decisions a year.
The claims document product aligns its governance to the state commissioners' model bulletin and the NIST AI Risk Management Framework through an independent vendor's control library. That alignment covers one product, and no bias testing result or fairness metric is published for any model.
The company authors the standard endorsements the industry uses to exclude generative AI liability from general liability cover, yet publishes no liability terms for the AI it sells. Filed loss costs and forms are public documents a consumer's counsel can obtain, but a policyholder whose claim a generated estimate shaped has no notice or route to contest.
Model Context Protocol connectors built with Anthropic bring the company's datasets into Claude workflows, and the company says a customer's AI strategy should not depend on one vendor's roadmap. Models beneath photo labeling, summarization and estimating features are unnamed, with no notice promised when one changes.
Estimating, assignment, contents and review products run as one connected claims stream, with generative features live inside four of them. Loss costs, forms and rules flow outward into carrier rating systems, but policy administration and billing belong to someone else.
Cloud delivery needs no special hardware and reaches most users within days of approval. Regions, processing location, residency commitments and the infrastructure provider are unpublished across operations in more than 20 countries.
Renewal is a price event, with increases taken at contract renewal alongside data and AI enhancements, per earnings commentary. Two AI estimating capabilities come at no extra cost with an existing professional license, though no rate card or billing unit is published.
Carriers of every size, reinsurers and brokers are served, along with restoration contractors in the estimating ecosystem and capital markets through catastrophe risk. Standard forms and filed loss costs are the base layer from which most American property and casualty carriers build their own rates.
Product updates
Material product, regulatory, evidence and commercial changes at Verisk, each verified against a live source and tagged to the capability axis it bears on. Funding rounds and awards are not product changes and are not logged.
Verisk and Equifax introduced Inflection Homeowners, a credit based insurance score built for property insurers that uses trended credit attributes instead of a point in time credit snapshot. It is aimed at homeowners underwriting, quoting and renewal decisions and extends the existing Inflection Auto model to property lines. Verisk states its own analyses show 2.6X lift between the lowest and highest risk bands.
Verisk released a US Data Center Exposure Database covering more than 2,500 facilities, available through Synergy Studio and Touchstone. It carries rooftop level geocoding, structural and operational detail such as construction type, floor area, capacity and redundancy, building footprint shapefiles and a 90 meter disaggregation grid, with Verisk noting available fields differ by facility and data source.
Alternatives to Verisk
Vendors in the same categories as Verisk whose profiles look most alike, each with a line on what the product does. Similar is not the same: check that each one does the job you need. No vendor pays for placement.
Akur8 is an actuarial platform for non-life insurance pricing and reserving, built on what the company calls Transparent Machine Learning.
OneShield sells core systems to property and casualty carriers and managing general agents.
EIS sells EIS OneSuite, a cloud native core platform for policy administration, billing, claims and customer management across property and casualty, life and annuities, group benefits…
Sixfold builds underwriting artificial intelligence for insurance carriers across property and casualty and life and health.
Insurity sells cloud based core software and analytics to property and casualty insurers, brokers and managing general agents.
Milliman is a global actuarial and consulting firm that also ships a portfolio of named, separately licensed analytics products to insurers.
The index publishes no overall score or ranking. See what we assess and the methodology.
Pricing
Vendor-published figures are labeled as such. Figures labeled “Estimated” are derived from third-party sources and have not been confirmed by the vendor.
| Entry Price | Pricing Basis | Data Protection Terms | Implementation | Source |
|---|---|---|---|---|
|
Not published. No rate card, unit price or contract term appears on any vendor surface, though two artificial intelligence estimating capabilities are stated to be included with an existing professional license at no additional purchase
|
Not published at unit level, though regulatory filings disclose the shape. Revenue is overwhelmingly subscription based, at 83 percent of second quarter 2026 revenue and growing 8 percent on an underlying basis, and the business reports across two groupings, underwriting covering forms, rules and loss cost services and extreme event solutions, and claims covering anti fraud and casualty solutions. That indicates recurring subscription contracts by product family rather than transactional or premium based charging, without establishing the unit within each. Some AI features are bundled into existing licenses and others are not, and nothing published distinguishes which is which beyond the two estimating capabilities named as included. | No tiered data protection terms are published, and the assurances that exist are structural rather than contractual. Data pooling occurs under licensed advisory organization and statistical agent status subject to periodic examination by state regulators, which places the aggregation of contributed carrier experience under public supervision rather than private terms, and a data stewardship function with published principles sits under corporate responsibility. One artificial intelligence commitment is specific and checkable: data retrieved through the protocol connectors remains within the client session and outside model training. A Commitment to Ethical and Responsible AI disclosure is published covering principles, governance structure, standards and review processes. Absent are a data processing agreement, retention schedule, subprocessor list, region list and any statement of what a carrier's contributed experience may be used for beyond the filed loss cost products. | No implementation, configuration or professional services fee is published, and the company's positioning is that implementation cost should be near zero because the artificial intelligence arrives inside products already in use. Published claims support that shape: the capability runs in the cloud and is embedded in platforms the customer already runs, no special hardware or software is required, onboarding is described as simple with no special setup, and most users can begin within days of approval. Two estimating capabilities are stated to require no additional steps or purchases for holders of an existing professional license. Set against that, the underwriting side involves filed forms, rules and loss costs integrated into carrier rating and policy issuance systems, which is substantial integration work at the carrier's own cost, and nothing published quantifies it. The comparison the company does publish is against building the capability internally, claiming that a carrier's own lawyers, actuaries and specialists could cost up to five times as much. | Regulatory Filing |
No rate card or unit price is published for any product. Because the company reports to the United States securities regulator, a good deal of verified commercial information is still public. Audited figures include second quarter 2026 revenue of $806.3 million, growth of 4.3 percent and 5.8 percent on an organic constant currency basis, subscriptions at 83 percent of revenue and rising, full year guidance of $3.19 to $3.24 billion, and adjusted margin guidance of 56 to 56.5 percent.
Those margins show a data franchise with pricing power rather than a services business, something a price list would not reveal. Earnings commentary describes increases taken at contract renewal alongside data and AI enhancements, so renewal is in effect a price negotiation.
How AI capability is charged where it is not bundled is unclear: some estimating features are stated to be included with an existing license, while nothing describes the commercial basis for the wider generative estate or the protocol connectors.