Definity to Embed Guidewire ProNavigator AI in Vyne Broker Platform

Definity Financial Corporation will bring Guidewire's ProNavigator AI assistant into its Vyne digital broker platform under a plan announced with Guidewire on November 12, 2026. The assistant, branded Definity Assist, will cover personal auto, personal property and Individually Rated Commercial Auto (IRCA — commercial vehicles priced per policyholder rather than by fleet) work.

Brokers will ask questions in everyday language and receive answers inside their existing Guidewire workflow, without switching applications. The companies presented the integration as a planned deployment. They did not provide an availability date.

The stakes are practical. Definity expects the assistant to make broker work more efficient, resolve inquiries faster and reduce the learning curve for new staff, while improving customer interactions. The project builds on Definity's existing policy-servicing workflows and on what the company describes as more than a decade of experience applying AI across insurance. For Guidewire, the deal extends ProNavigator, its in-application assistant, into a major Canadian insurer's broker channel.

What Will Definity Assist Actually Do?

The core function is question answering in context. A broker working on a policy in Vyne will be able to request information in plain language and see the answer within the current workflow. Definity frames the benefit in three areas: efficiency, faster inquiry resolution and a shorter ramp-up period for new brokers.

The scope is bounded. Three lines of business are named — personal auto, personal property and Individually Rated Commercial Auto. The announcement does not specify Definity's exact configuration of the product, so the feature set depends on what the insurer chooses to switch on.

The deployment also gives Definity a second, less visible function. Broker queries are intended to show where help content, training and policy documentation need improvement, and the resulting insights will inform process changes and future AI-driven broker experiences. Individual brokers get help in the moment. Patterns across their questions can expose recurring weaknesses in supporting material. Acting on those patterns would allow Definity to address the source of repeated inquiries as well as the inquiries themselves — a feedback loop between usage and documentation.

Which Models and Architecture Power ProNavigator?

Guidewire's ProNavigator documentation places the assistant in a sidebar within core applications, where it can supply answers, suggestions and actions based on the user's current work. The product runs on Anthropic's Claude models. Guidewire uses the Converse API (an interface for sending documents to Claude for interpretation) to read tables and graphics in documents such as underwriting manuals, and Claude's Citations feature to attach paragraph-level references to answers.

For retrieval-augmented generation (RAG — a technique that grounds model answers in retrieved source documents rather than model memory), each customer receives its own vector database. A vector database stores text as numerical embeddings so semantically similar passages can be found quickly. Guidewire says the dedicated store keeps client data segregated from other customers.

Access control works at the level of business categories. ProNavigator maps access rules to categories such as region or insurance line and uses single sign-on to apply role-based permissions consistently across source documents, searches and generated responses. A broker who cannot open a document should not see its contents surface in an answer either.

How Are Compliance and Data Protection Handled?

ProNavigator records queries, responses and their citations for compliance teams, audits and claims disputes. That audit trail is the mechanism by which an insurer can reconstruct what an assistant told a broker and on what source paragraph the answer rested.

Guidewire says human oversight applies to every AI-generated insight. The architecture, according to the company, protects personally identifiable information and keeps proprietary customer data from being used to train public models. Those commitments matter in insurance, where policyholder data is subject to privacy regulation and where underwriting manuals and rate content are competitive assets.

Two caveats from the announcement itself. These architecture details describe ProNavigator generally. The announcement does not specify Definity's exact configuration. And the deployment remains a plan, without a release date.

Why Does This Deployment Matter for Insurance AI?

The announcement tests a proposition the insurance industry has debated since generative models entered enterprise workflows: whether faster answers can coexist with source citations, access controls and human review. ProNavigator's design attempts to reconcile the two. Answers carry paragraph-level citations. Permissions follow the user. Oversight is logged.

The second test is operational. Definity intends to turn broker query data into better supporting material — using the assistant not only to answer questions but to reveal which questions keep coming back. If the insurer acts on those patterns, the deployment becomes an instrument for fixing documentation and process, not just for deflecting inquiries.

The immediate audience is Definity's broker network. The broader audience is other insurers weighing similar deployments, for whom the project will show whether an in-workflow assistant on Claude-class models can clear the compliance bar that insurance regulators and audit teams apply. No date has been set, but the November 12, 2026 announcement commits both companies to a configuration whose details — model versions, document scope, citation behavior in production — will determine whether brokers trust the answers it gives.