How Unipol Uses IBM watsonx to Cut IT Response Time by 97%

Unipol Assicurazioni has spent more than six decades building Italy’s most extensive insurance network. With 16.8 million customers, 12,000 employees, and operations spanning auto, home, health, life, and asset management, the company’s IT infrastructure underpins an enormous volume of daily transactions. As the organization pursued enterprise-wide AI adoption, it recognized that its IT operations layer needed to be rebuilt from the ground up.

The status quo left significant gaps. Before the transformation, Unipol’s control room monitored only 26% of system events in near real-time, and when incidents occurred, the average response time was 20 minutes. Accounting and claims processing cycles ran to 21 hours. With fragmented monitoring tools and no unified intelligence layer, technical teams spent the majority of their time on incident triage rather than strategic work.

Unipol and IBM co-developed NAMI (Next Automation Monitoring Insurance), an AI platform built on the IBM watsonx ecosystem. IBM watsonx.ai provided the generative AI foundation with natural language interfaces for instant operational insights. IBM watsonx Orchestrate managed agentic workflows and automated decision routing. IBM Cloud Pak for Data handled data preparation, while IBM Cloud Pak for AIOps unified event management across the entire application estate. The platform integrates over a dozen monitoring systems into a single data lake. NAMI uses a multi-model approach, combining IBM Granite foundation models, Meta’s Llama, and Mistral’s Pixtral AI to match the right model to each task. The initial deployment ran on IBM Cloud before migrating to a hybrid architecture using Red Hat OpenShift for deployment flexibility and IBM Fusion HCI for on-premise governance.

The results arrived quickly. In its first two months, NAMI analyzed more than 800 system events, autonomously resolving the majority and escalating only 538 for human review. Monitoring coverage jumped from 26% to 100%, and the average event response time collapsed from 20 minutes to 90 seconds—a 97% reduction. Incident handling time fell 90%, freeing technical teams to focus on higher-value work. Accounting and claims processing cycles shortened from 21 to 18 hours.

Unipol is now extending NAMI toward full event automation, AI-driven process transformation, and support for IBM z17 mainframe infrastructure. The platform’s architecture—agentic workflows, multi-model intelligence, and unified observability—provides a foundation for continuous optimization across an enterprise that touches nearly one in four Italians.