Verisk pulls data into Claude

Insurance AI has had a basic problem for a while — the model can sound smart, but the real evidence still lives somewhere else. Underwriters still have to pull ISO filings. Claims teams still have to open estimating tools. So the work turns into copy-paste, tab switching, and a lot of “go check the source.” Verisk’s move this week is about closing that gap by putting its insurance data and tools directly inside Anthropic’s Claude through Model Context Protocol, or MCP, connectors. ### What actually launched? On May 5, Verisk said its insurance analytics are now available in Claude through standardized Verisk MCP connectors, with the pitch aimed at insurance and property restoration professionals working inside secure enterprise AI environments. Verisk framed the launch as a way to let people ask for insights in natural language while keeping governance and human decision-making in place. ### What is MCP doing here? MCP is basically the plumbing that lets a model reach into outside systems in a structured way instead of guessing from its training data. Anthropic has been pushing this connector model hard across financial services, with new partner integrations and an MCP app meant to let Claude work against the data professionals already use. Verisk is now one of those insurance-specific data layers. ### Which Verisk data shows up first? The first concrete pieces are pretty telling. Verisk’s Underwriting Intelligence connector lets Claude pull current data from ISO’s loss-cost system, including filed loss-cost changes, impact summaries, loss experience trends, pure premium movement, and preliminary indications. That means an underwriter can ask Claude what changed in a filing and why, instead of hunting through separate systems. ### What about claims and restoration? Verisk also published a connector for XactRestore and XactRemodel. That one lets Claude turn a natural-language loss description into a structured estimate with rooms, scope, and line items, and even generate multiple estimate tiers for comparison. So this is not just “ask a chatbot about insurance.” It reaches into an actual estimating workflow where speed and consistency matter. ### Why does that matter operationally? Because insurance work usually breaks at the handoff between judgment and evidence. A model can draft a recommendation, but if the user still has to leave the workflow to verify pricing, filings, or scope, the time savings collapse. Verisk is trying to make Claude useful at the exact moment of decision — where an underwriter is pricing risk or that is the real product idea here. ### Is this just a chatbot wrapper on old data? Not really — the more important change is where the data shows up. Verisk already sells the underlying analytics. What changed is the interface and the workflow position. Instead of training people on another dashboard, the company is betting that natural-language access, which is exactly how Verisk is describing them. ### Why now? Anthropic has been building out Claude for regulated professional work, especially in financial services, with connectors, Microsoft 365 integrations, and agent-style tooling. Verisk’s launch fits that bigger shift. The market is moving from general-purpose AI demos to domain systems that can reach trusted internal or licensed data. Insurance is a natural fit because the work is document-heavy, rules-heavy, and expensive when wrong. ### Bottom line The headline is not that Verisk “added AI.” It already had the data. The headline is that Verisk is trying to put insurance-grade evidence inside the chat interface where underwriting and claims decisions increasingly start. If that sticks, Claude becomes less of a drafting tool and more of a front end for actual insurance work.