Zurich Insurance Group

Global property and casualty insurers process enormous volumes of customer submissions daily — risk descriptions, loss reports, site photographs, and broker correspondence — arriving in dozens of languages and formats. For Zurich Insurance Group, one of the world's largest P&C carriers operating across 170 countries, this unstructured, multimodal data created a fundamental mismatch with traditional rules-based underwriting systems designed for structured inputs. Translating handwritten inspection notes, multilingual emails, and photographic evidence into actionable risk assessments was slow, inconsistent, and heavily dependent on individual underwriter expertise. The result: longer turnaround times, uneven risk evaluation quality, and a growing backlog that constrained Zurich's ability to serve customers at the speed modern commercial clients expect.

Zurich partnered with Microsoft to deploy Azure OpenAI Service as the foundation for a suite of bespoke, multimodal AI applications. Rather than replacing existing workflows wholesale, the program followed an incremental build-and-extend model — beginning with targeted underwriting tools and expanding across the full insurance value chain. The platform ingests unstructured inputs — scanned documents, images, emails, and multilingual reports — and surfaces structured risk insights for underwriters. Critically, the AI operates as a decision-support layer rather than an autonomous decision-maker, helping underwriters ask sharper questions and synthesize complex risk profiles faster. Azure's enterprise-grade security and compliance architecture allowed Zurich to deploy consistently across its decentralized global operations. To date, more than 200 AI tools have been built and deployed on the platform.

The rollout has produced measurable improvements in underwriting speed, accuracy, and customer experience across Zurich's global operations:

• 200+ AI tools deployed enterprise-wide, spanning underwriting and adjacent value chain functions
• Faster risk evaluation turnaround times, reducing delays for commercial clients and brokers
• More consistent risk assessments across geographies and underwriter experience levels
• Underwriters report improved ability to identify critical risk factors and engage customers with more targeted questions

Beyond efficiency gains, the program has shifted how underwriting expertise accumulates within the organization — newer underwriters ramp faster by working alongside AI-assisted workflows, compressing the learning curve that has historically defined P&C talent development.

• Multimodal AI unlocks global scale: Rules-based systems cannot handle the linguistic and format diversity of real-world P&C submissions — generative AI is essential for unstructured, multilingual data at scale.
• Coach, don't replace: Positioning AI as an underwriter's coach — surfacing insights rather than making decisions — drives adoption and preserves the human judgment that regulators and clients expect.
• Enterprise infrastructure is a prerequisite: Deploying 200+ tools across a decentralized global operation requires a secure, compliant cloud foundation — selecting a partner with enterprise-grade infrastructure early prevents rework.
• Expand incrementally: Starting with high-impact underwriting use cases and systematically extending across the value chain is more sustainable than attempting a single large transformation.