Building a central data, analytics, and AI solution using AWS with Generali

Generali consists of more than 40 operating entities, and all teams rely on similar large-scale processes, such as claims and underwriting. The company wanted to centralize and standardize data access for these shared workflows to improve efficiency and accuracy in data management. So, Generali needed a solution to consolidate a common data view and develop and deploy common AI applications or artifacts across the company without duplicating efforts. “Our objective was to scale up adoption while maintaining accuracy and sophistication everywhere,” says Andrea Pietrasanta, group head of data, AI, and automation at Generali.

Generali first collected data from group sources and operating entities. It then migrated its existing analytics solutions from third-party providers to AWS to create a central data and analytics solution. Finally, it built new AI applications on top. “We received constant, ongoing support from AWS to make the most of services and technology and enhance our team’s expertise,” says Antonio Montuschi, head of group AI and generative AI development at Generali.

Generali built its central data and analytics solution on Amazon Redshift, which organizations use to power data-driven decisions with excellent-price-performance cloud data warehouse. The solution ingests text, images, and structured data from both local operating entities and external sources, such as weather data and satellite images. Using AWS Glue—a serverless service for discovering, preparing, and integrating data at virtually any scale—the company can process data through data-quality and transformation pipelines and generate insights. Generali established rules and policies for its cloud infrastructure to comply with internal policies and with regulations such as the General Data Protection Regulation, which provides guidelines on client data privacy.

To automate data processing and decision-making, Generali uses Amazon SageMaker AI—a service for building, training, and deploying machine learning models. For example, the company can automate claim settlements and use predictive modeling to propose the appropriate payout in an injury claim negotiation. Generali developed 16 flagship use cases that are related to pricing, underwriting, claim processing, and operations—achieving high impact, accuracy, and scalability. Different operating entities at Generali can quickly implement these use cases by using the company’s centralized Global AI engine.

Using Amazon Bedrock, Generali also enhanced the customer experience with a generative AI assistant that provides automated, instant responses to assistance requests all day. Previously, customers would send email messages and wait for manual replies. Using AI, Generali can also accelerate claim settlements and quotations. “When automating a service, it becomes more standard and consistent and less prone to error,” says Pietrasanta.

In the first 3 years of implementing its flagship use cases, Generali saved more than EUR €200 million in operating costs by using AI and animation. It has also improved technical results—for example, by evaluating eligibility and claim amounts more accurately. The company has accelerated its customer response time by automatically processing more than 21 million API calls per month. It has decreased claim settlement time from several days to 1 day or, for simple health claims, to seconds. In addition, Generali has rapidly expanded its use of automation since implementing its central analytics solution on AWS. The company increased the number of its AI applications from 5 to over 50 in 3 years, with a goal of 200 applications in the next 3 years.

Generali aims to scale up the adoption of its flagship use cases across different operating entities and continue investing in research and innovation. For example, Generali is developing several generative AI use cases and advanced risk modeling. “Using AWS tools and functionality, we can customize the architecture to build the solution we want in terms of features and quality,” says Pietrasanta. “This keeps costs lower than adopting packaged solutions.”