AXA Switzerland's data and AI work on Google Cloud has reduced query and processing times by over 95%, from days to minutes or seconds, and its Smart Data initiatives have generated a high double-digit million Swiss franc profit improvement. Over 100 Google Cloud projects are now live in production across all business lines. “Our human specialists used to spend 25 minutes reading complex medical documents to extract key information,” says Amit Jain, Head of Artificial Intelligence and Data Innovation. “Now AI completes this in seconds with high accuracy, allowing our experts to handle more cases and focus on complex decisions.”
“Our Smart Data projects were among the most profitable initiatives in AXA Switzerland's history,” said Gerhard Dolge, Chief Data and AI Officer, AXA Switzerland. “A case in point: by implementing a simple, existing fraud detection ruleset, we uncovered a 120,000 Swiss franc false claim involving intoxicated workers—in its very first week after deployment.”
In 2021 the insurer faced a choice between traditional insurance models and a digital-first future. Its legacy infrastructure had become a barrier: complex queries could take an entire weekend, and siloed systems hampered cross-functional insights. “Our data structures were designed for reporting, not real-time decision making,” Jain said.
After a proof of concept, AXA Switzerland chose Google Cloud's multi-cloud approach and in 2021 migrated its existing data lake to Google Cloud. AXA Switzerland was the first subsidiary in the AXA Group to adopt cloud technology fully. It then applied machine learning and deep learning to optimise operations, experimenting with voice bots and document analysis, and it is now integrating generative AI while focusing on responsible use.
BigQuery brings CRM, ERP, ecommerce and operational data into one system; Vertex AI runs machine learning workflows; Cloud Run handles serverless application deployment; and Dialogflow supports voice interfaces and chatbots. For generative AI, the company favours assistant systems that augment employees over direct customer interactions, puts every AI initiative through an internal ethics and risk assessment, and focuses on natural language processing such as translating and summarising court rulings and medical records. It is also implementing automated triage to determine which claims require expert evaluation, to make the best use of its vehicle experts, and it reorganised IT by merging its operations and cloud competence teams.
Next, AXA Switzerland is exploring voice bots and agent assist for call center and claims staff, multimodal AI across text, voice, images and structured data, satellite data for underwriting, claims management and prevention, and agentic AI for specific insurance workflows.