Zurich
genAI call analysis and real-time agent guidance for retention
Voice IQ, Zurich's genAI solution, analyzes retention calls and guides agents in real time; deployed at Sabadell Seguros in Spain, it raised retention by 20% and received a Qorus Innovation in Insurance 2025 award.
Objective
Reduce cancellations by equipping agents during the call: detect the reason, propose the right response, and personalize the retention offer at the right moment.
The deployment
Voice IQ analyzes thousands of client calls to give contact center teams real-time guidance and recommendations tailored to each conversation. In practice, the tool listens to retention exchanges, identifies what is at stake, and suggests the most relevant response and offer to the agent to prevent cancellation. Zurich first deployed the tool at Sabadell Seguros, its joint venture in Spain, where retention rose by 20%. The solution received a Qorus Innovation in Insurance 2025 award (GenAI Innovation of the Year category, bronze), described as a genAI platform that analyzes retention calls to strengthen the agent-client relationship, improve service practices, and reduce churn. Voice IQ is part of a wider portfolio of Zurich AI tools on the underwriting and service side, including Guideline IQ, an assistant for underwriters already in use across 29 entities.
Results Proof C
The +20% retention result was published by Zurich on its official AI page (interested party), corroborated by a third party through the Qorus / NTT DATA 2025 award that documents the VoiceIQ solution by name at Sabadell Seguros. A self-reported figure from the brand, not an audited financial result.
How it works
Inferred typical approachThe internal detail is not public. Here is a proven approach that leads to the same result, to adapt to your stack.
The stack in detail
- outil Voice IQ Solution genAI developpee par Zurich : analyse des appels de retention et guidage temps reel du teleconseiller (reponse et offre recommandees).
- llm LLM de Voice IQ (modele exact non publie) Zurich decrit une plateforme genAI sans nommer le modele sous-jacent.
- infra Transcription et speech analytics des appels Conversion des conversations en texte exploitable ; sans transcription fiable et retour sur l'issue, les recommandations cessent de s'ameliorer.
How it runs, concretely
For ops teams-
1Call analysis AI
Voice IQ analyzes thousands of calls to spot cancellation reasons and best practices.
-
2Real-time guidance AI
During the call, the tool suggests the fitting response and retention offer to the agent.
-
3Agent action agent
The agent applies or adjusts the recommendation and leads the conversation with the client.
The content of the retention conversations and the outcome (cancellation avoided or not). Without reliable transcription and feedback on the outcome, the recommendations stop improving.
How your customers perceive this type of use
Sourced studiesLes consommateurs n'acceptent pas les chatbots par defaut : 64% prefereraient que les entreprises n'utilisent pas d'IA dans leur service client (Gartner, 2024) et pres d'un utilisateur sur cinq du service client par IA n'en retire aucun benefice (Qualtrics, 2025). L'acceptation se construit sur trois conditions mesurees par Salesforce : savoir qu'on parle a une IA, pouvoir escalader vers un humain, comprendre la logique de l'agent.
Acceptance conditions
- Etre informe qu'on parle a une IA et non a un humain (pres de 75% le demandent, Salesforce 2024)
- Un chemin d'escalade clair vers un agent humain (45% plus enclins a utiliser l'agent IA, Salesforce 2024)
- Une logique de l'agent clairement expliquee (44% plus enclins, Salesforce 2024)
Red lines
- Rendre l'humain injoignable : c'est la premiere inquietude des consommateurs sur l'IA dans le service client (Gartner 2024) et 50% craignent que l'IA les coupe du contact humain (Qualtrics 2025)
- Remplacer le service client par l'IA sans alternative : 53% envisageraient de partir chez un concurrent (Gartner 2024)
Sources: Salesforce 2024 · Gartner 2024 · Qualtrics 2025
How to replicate
Inference, not sourcedData prerequisites
- Recordings and transcripts of retention calls
- Labeling of outcomes (cancellation avoided or not)
- Catalog of retention offers
Org prerequisites
- Contact center with trained agents
- Legal and social framework for call analysis
- Quality loop to validate the recommendations
Possible stack
- Custom genAI solution (the Voice IQ route)
- Speech analytics platform plus a market real-time assistant
- CCaaS with an agent-assist module
The plan, step by step
- Step 1Frame the legal side (informing clients and agents, works councils) and secure access to the recordings.Deliverable: Validated framework plus an accessible corpus of retention calls.
- Step 2Analyze the cancellation corpus: recurring reasons and the practices of the best agents.Deliverable: Taxonomy of reasons plus a playbook of responses by reason.
- Step 3Build the guidance (recommendations by reason, associated retention offers).Deliverable: Prototype tested on recorded calls, outside production.
- Step 4Launch the real-time pilot on one team against a control team.Deliverable: Read on retention and call quality, pilot versus control.
- Step 5Close the learning loop on call outcomes and plan the extension.Deliverable: Refined recommendations plus a rollout plan to the other teams.
First step: Analyze a corpus of cancellation calls to identify recurring reasons before connecting real-time guidance.
The same pattern, proven in other industries
ADT
genAI customer agent
Air India
genAI customer agent (automated customer service)
AT&T
Governed enterprise genAI assistant: an LLM connected to internal knowledge, open to tens of thousands of employees and called at scale by automated systems
Sources
- S1 AI at Zurich - Zurich Insurance Primary archive pending
- S2 Qorus and NTT DATA announce winners of the tenth Innovation in Insurance Awards Established press archive pending
An error, newer info, a source?
This page lives on its accuracy. If a figure has moved, if the deployment has changed, or if you have a higher-quality source, tell us. Every sourced correction is verified before publication.