Claude AI + Xactimate: What Verisk’s Anthropic Partnership Really Means for Restoration Contractors

On May 5, 2026, Verisk, the parent company of Xactware, Xactimate, and XactAnalysis, announced that its insurance analytics are now available inside Anthropic’s Claude via standardized Model Context Protocol (MCP) connectors. Industry headlines lit up immediately: “Claude writes Xactimate estimates,” “AI cuts two hours off every estimate,” “The end of manual scoping.”

At WrightWay, we build and review hundreds of Xactimate estimates a year for water, fire, mold, and storm losses across Southwest Florida. We dug into the announcement, the technical documentation, and the early third-party analysis so you don’t have to. The reality is more nuanced than the headlines, and for most restoration contractors, the practical impact today is smaller than the press release suggests.

This is what was actually announced, what it does, what it doesn’t, and how restoration shops should think about it.

What Verisk actually launched

Per the official Verisk press release, Verisk built two Claude MCP connectors:

• Verisk Underwriting Intelligence (ISO Indications): for insurance carriers and actuaries. Conversational access to loss-cost trends, experience insights, and ISO filing signals. Reinsurance News coverage has additional detail.
• Verisk XactRestore: for restoration contractors. Conversational access to pricing and estimating intelligence powered by Verisk’s zip-code-specific pricing database.

The restoration-facing piece is the one that matters to us. And right there in the name is the wrinkle that most headlines glossed over: the connector is for XactRestore, not Xactimate.

Xactimate vs. XactRestore: the distinction every estimator needs to know

Verisk sells three different estimating products that share the same underlying Xactware price book but ship as separate applications (C&R Magazine breaks the product distinctions down here):

• Xactimate: the industry standard. The platform virtually every carrier, TPA, and large restoration contractor uses to write the estimate that ultimately goes to the insurance company. Sketches, line items, ESX files, and carrier-facing approvals all live here. The Pro tier runs roughly $250 to $300 per user per month.
• XactRestore: a newer Verisk product positioned at restoration contractors who want job-management features bundled with estimating. Different interface, different conventions, same underlying price data. Pricing isn’t published; you have to request a quote.
• XactRemodel: the less-discussed sibling, aimed at remodelers doing non-insurance work. Runs around $105 to $129 per user per month.

The Claude connector lives inside XactRestore only. If your team writes carrier-facing estimates in Xactimate, and almost every restoration company does, the connector does not, today, draft those estimates directly. It drafts them inside XactRestore, and XactRestore’s Pro tier offers a path to import that work into Xactimate.

That’s an important distinction the press cycle largely missed. “AI writes your Xactimate estimate” and “AI writes a draft inside a different Verisk product that you then import into Xactimate” are two very different value propositions for a busy estimator.

What the workflow actually looks like

Based on Verisk’s published documentation and the first independent walkthrough at FindSkill.ai, here is the workflow a restoration contractor follows today:

1. Prerequisites: a Claude Pro or Team subscription (about $20 per user per month), an existing Verisk pricing-data entitlement, an XactAnalysis profile, and the XactRestore Claude connector provisioned by your Verisk account manager. There is no self-serve sign-up at launch.
2. Input: drop job-site photos into Claude along with a short memo with client name, property address (the zip code is critical, since Xactware pricing is zip-specific), damage category (water, fire, mold), affected square footage, structural type, and the carrier or TPA involved.
3. Output: Claude returns a structured estimate with line items in Xactimate format, quantities estimated from the photos, and zip-coded Verisk pricing already applied. Independent reviewers describe the output as roughly 80 to 90 percent of a usable estimate.
4. Human review: the contractor adds the line items that are almost always missing, including pre-1980 asbestos and lead testing, tenant relocation and Additional Living Expense flags, mold tier escalations as Category 3 water transitions to Tier 1 or 2 mold within 48 to 72 hours, and the emergency-extraction time-and-materials vs. unit-price variance some carriers require for the first 48 hours.
5. Export: the cleaned-up estimate exports to standard Xactimate format and continues through XactAnalysis to the carrier exactly the way it does today.

Verisk’s claimed time savings

Verisk publicly claims experienced contractors can save 30 minutes to two hours per estimate. At a loaded labor rate of $80 to $120 per hour, the back-of-the-envelope math runs from roughly $10,000 of annual savings per estimator at five estimates per week up to $80,000+ for high-volume CAT response shops doing ten or more estimates per week.

Those numbers are real if, and only if, the conditions are right. Independent walkthroughs note the learning curve to stabilize savings takes 8 to 12 estimates per estimator, and the workflow is most useful for shops writing at least three jobs per week. Below that volume, you spend more time learning XactRestore than you save in Claude.

The honest limitations

Drawing from Verisk’s own documentation and independent industry coverage, here is what the Claude and XactRestore connector cannot do today:

• It does not write estimates directly inside Xactimate.
• It does not work outside the United States and Canada, where Verisk pricing data is most granular.
• It does not adjudicate scope disputes with adjusters.
• It does not replace your XactAnalysis or carrier-portal workflow.
• It does not comply, on its own, with carrier policies that prohibit AI-prepared estimates, and those carrier policies are still inconsistent across the industry.
• It does not address state contractor-licensing guidance on AI-generated work product, which has not yet been published in most jurisdictions.

Governance and data trust

Verisk emphasizes that the connectors run within their existing governance framework: regulated-grade proprietary data, a governed analytics layer on top of trusted pricing, explainable outputs with human accountability, and a model-agnostic posture (this is Claude today; nothing technically prevents the same connectors from supporting other LLMs later). Reinsurance News and FinTech Global both cover the governance positioning in depth. For restoration contractors who work with carriers that require auditable estimating, this matters, because the data inputs are the same Verisk pricing data carriers already trust.

Verisk CEO Lee Shavel framed it this way in the announcement: “Trust is the foundation of insurance, and that doesn’t change as new technologies emerge. Responsible use of generative AI isn’t just about efficiency. It’s about helping insurers make sound decisions.”

How WrightWay is thinking about it

We are not rushing to bolt this onto our daily Xactimate workflow, and we don’t think most restoration shops should either. Here’s our internal take:

• If your team lives in Xactimate, the connector doesn’t yet remove keystrokes from the estimate that goes to the carrier. The savings depend on your team learning XactRestore.
• The strategic signal matters more than the launch product. Verisk chose to push the conversational front end to Claude specifically. If a Xactimate-direct connector follows in 6 to 12 months, which we think is likely, the restoration companies that already have Claude workflows built will move faster than those that don’t.
• A real pilot beats theory. If you want to evaluate this honestly, pick one estimator, one clean job type (Cat-2 water is the easiest), and run 8 to 12 jobs end to end through XactRestore plus Claude. Measure actual minutes saved versus carrier rework. That tells you the truth for your shop.
• Carrier policy is the real unknown. Until carriers publish clear guidance on AI-assisted estimates, the contractor still owns the work product. Build review steps into your workflow regardless of how good the draft is.

The takeaway

Verisk’s Claude integration is real, technically interesting, and a credible signal about where insurance estimating is heading over the next two to three years. It is not, today, a drop-in upgrade for restoration companies running their carrier-facing work in Xactimate. The smart move for most shops is to track it closely, run a small honest pilot if you have the bandwidth, and stay ready for the next wave, which will almost certainly include a more direct Xactimate path.

If you’re a Southwest Florida property owner working with WrightWay on a water, fire, mold, or storm claim: the estimate you sign off on still comes from a licensed, experienced WrightWay estimator who reviews every line. Tools change. Accountability doesn’t.

Sources and further reading

• Verisk press release: “Verisk Brings Its Trusted Analytics and Generative AI Capabilities Directly into Anthropic’s Claude” (May 5, 2026)
• Anthropic: Model Context Protocol overview
• Reinsurance News: Verisk integrates insurance analytics and gen AI with Claude via MCP connectors
• C&R Magazine: Claude and Xactimate: The Headline vs. The Reality
• FindSkill.ai: Verisk XactRestore in Claude, a 30-min walkthrough
• FinTech Global: Verisk integrates data tools into Claude for insurers
• Stocktitan: VRSK investor coverage of the Claude integration