AXA Switzerland

AXA Switzerland, the country's leading insurer serving over 40% of Swiss companies, faced a data infrastructure crisis that was actively constraining its ability to compete. Legacy systems built for batch reporting meant complex analytical queries could consume an entire weekend of processing time, making real-time underwriting decisions and dynamic pricing effectively impossible. Siloed CRM, ERP, and operational data prevented cross-functional insights critical to accurate claims management and fraud detection in property and casualty lines. As Amit Jain, Head of AI and Data Innovation, noted, the architecture was designed for reporting, not decision-making — a structural liability in an industry where speed and accuracy directly determine profitability and loss ratios.

In 2021, AXA Switzerland migrated its data lake to Google Cloud, consolidating CRM, ERP, ecommerce, and operational data into BigQuery as a unified analytics foundation — eliminating the silos that had blocked cross-functional insight. On top of this, the company deployed Vertex AI for machine learning and deep learning workloads, and Cloud Run for serverless application deployment. The current phase layers Generative AI (via Google's Gemini models) for document processing: AI systems now summarize complex medical records and court rulings to support claims specialists. Critically, AXA adopted an assistant-first deployment model — generative AI augments human specialists rather than replacing them or facing customers directly, a deliberate architectural choice to manage hallucination risk under Swiss regulatory constraints.

The transformation delivered outcomes that AXA Switzerland's Chief Data and AI Officer Gerhard Dolge described as among the most profitable initiatives in the company's history:

• 95%+ reduction in query and processing times — from days (sometimes a full weekend) to minutes or seconds
• High double-digit million CHF profit improvement from Smart Data initiatives across pricing, damage expense management, and marketing
• Medical document review time cut from 25 minutes to seconds, enabling claims specialists to process significantly more cases
• Over 100 Google Cloud projects live in production across all business lines
• A fraud detection ruleset uncovered a 120,000 CHF false claim in its first week of deployment

Organizationally, IT shifted from a service-delivery function to an innovation partner driving proposals to the business.

• Sequence matters: establishing a unified cloud-native data foundation (BigQuery consolidating all enterprise sources) was a prerequisite — attempting to layer AI on fragmented legacy systems would have failed.
• Assistant-first is the right entry point for regulated insurers: deploying generative AI to augment specialists rather than automate customer interactions manages compliance and hallucination risk while still capturing material productivity gains.
• Organizational restructuring is non-negotiable: merging separate IT operations and cloud competence teams was as important as the technology itself in sustaining innovation velocity.
• Start with internal, high-volume document workflows: medical records and claims documents offer clear ROI with measurable before/after baselines and lower regulatory exposure than customer-facing automation.