Fastweb + Vodafone details two LangGraph-based AI agents for customer support in Italy
A LangChain case study published on December 16 outlines how Fastweb + Vodafone is using LangGraph to run agent workflows for customer-facing support and call-center assistance. The write-up includes metrics like 90% correctness for Super TOBi and 86%+ one-call resolution for Super Agent.
LangChain published a case study on Fastweb + Vodafone (Swisscom Group) on December 16, describing two LangGraph-based AI-agent initiatives for customer support in Italy.
The customer-facing project, Super TOBi, extends the existing TOBi chatbot with a Supervisor graph that applies guardrails and routes queries to specialized Use Case graphs, including clarification questions when routing is uncertain.
Those Use Case agents follow an LLM Compiler pattern, call a limited set of customer APIs, and can return structured action tags that trigger transactions like offer activation, service disablement, or payment method updates inside the chat interface.
LangChain reports Super TOBi currently serves nearly 9.5 million customers via the Customer Companion App and voice channels, covering use cases such as cost control, roaming, billing, and sales.
The published metrics for Super TOBi include a 90% correctness rate, an 82% resolution rate, and a Customer Effort Score of 5.2 out of 7.
Fastweb + Vodafone’s internal tool, Super Agent, uses LangGraph orchestration with LangChain tools and stores operational procedures as a Neo4j knowledge graph rather than answering customers directly.
In the same write-up, Fastweb + Vodafone attributes One-Call Resolution rates above 86% to this consultant-assist approach, and describes an automated ETL pipeline plus CI/CD that can promote updated procedures to production within hours without downtime.