Chatbot Agenti: Unipol's innovative assistance with the power of GenAI
Cluster Reply, in collaboration with Sprint Reply, supported Unipol in enhancing its agency network support through a virtual assistant powered by Generative AI.
#GenAI
#Microsoft Azure
#RAG
THE CHALLENGE
Reduce the inflow to Assisted Help Desk systems and simplify the management of transactional requests.
THE CONTEXT
Unipol is a leading multi-line insurance company in Italy, with a widespread presence throughout the territory via an extensive network of agencies and sub-agencies. With a comprehensive offering of innovative insurance solutions, the company has always focused on digital evolution to improve its processes and the services offered to its network of agents.
The increase in the volume of requests handled by the agency network and the need to access business information quickly and reliably makes it necessary to simplify the consultation of the knowledge base and lighten the load of tickets directed to Help Desk systems (AHD), while simultaneously improving the management of operational activities.
THE SOLUTION
Generative intelligence that improves service quality and reduces manual workload
Cluster Reply and Sprint Reply developed Agent Chatbot, a solution based on Generative AI designed to effectively support the network of agents and operators.
The chatbot offers dual functionality: it facilitates immediate consultation of corporate knowledge, integrating structured sources and external services, and automates the execution of dispositive requests on internal systems.
Fully integrated with Unipol's technological ecosystem, Agent Chatbot allows for a reduction in ticket submissions to AHD systems, improving the user experience and increasing the overall efficiency of agency activities.
)
THE METHOD
RAG architecture ntegrating semantic search and generative AI
To develop the solution, Cluster Reply and Sprint Reply adopted an architecture based on Retrieval-Augmented Generation (RAG), combining Azure AI Search and Azure OpenAIto dynamically query the corporate knowledge base.
The system is built around a network of specialized agents, orchestrated through LangGraph and coordinated by LangChain, which manages conversations, tool activation, and document ingestion. Each agent performs specific functions, such as retrieving information, validating content, and interacting with corporate systems.
The platform includes a module for document processing, allowing the upload and analysis of files using multimodal models like GPT-4o, integrated with Microsoft OCR services for content extraction.
Integration via REST API with business systems and external tools enables the automation of disposal requests. The entire platform, deployed on the Azure cloud environment, ensures scalability, security, and resilience.
THE RESULTS
Faster, more efficient, and automated request management
Reduction of 100,000 informational tickets annually
Decisive decrease in AHD inflow.
65,000 automated operations every year
Automation of dispositive requests.
Optimization of management times
Daily requests handled more quickly.
Increase in operational efficiency
Greater precision in the daily management of activities.
Unipol is one of the leading insurance companies in Italy, with over 3,200 employees and a widespread network of 2,000 agencies across the national territory. The company offers a wide range of innovative products for individuals and businesses, combining expertise, reliability, and customer proximity. Guided by values such as transparency, inclusion, and social responsibility, Unipol promotes solutions focused on well-being and growth, offering favorable conditions to its employees and partners, with the aim of creating simple and effective tools that meet the real needs of its customers.