International Financial Services Firm (unnamed)

A large international financial services firm operating across 11 countries faced mounting pressure on its customer support infrastructure. Fragmented call centers and service portals made it difficult for policyholders to consistently reach accurate information — a critical failure point in Property & Casualty, where customers often contact support during stressful claim events. As inquiry volumes scaled to hundreds of thousands annually, the existing model could not deliver consistent service quality without proportional increases in staffing costs. The gap between customer expectations for immediate, accurate responses and the firm's operational capacity was widening, creating both a customer satisfaction problem and a cost sustainability problem.

CGI designed and implemented an AI-driven conversational chatbot powered by Natural Language Processing (NLP) to automate the end-to-end customer care process across the firm's multi-country operations. The solution was built using a DevOps delivery model, enabling iterative development and continuous improvement rather than a single-phase rollout. CGI brought a multidisciplinary team covering linguistics, NLP database construction, user experience design, and cloud and data integration — addressing the firm's gap in implementation expertise despite having a clear technology direction. The chatbot was integrated with existing service portals and data systems to enable accurate, context-aware responses. The engagement was structured as an ongoing advisory relationship, positioning the solution to evolve alongside advances in LLM technology and next-generation conversational AI.

The chatbot now handles 500,000 customer support conversations per year, delivering measurable operational and financial impact:

• €2 million in annual cost savings, with savings growing year over year
• 50%+ of inquiries resolved without human intervention, eliminating the need for agent involvement in the majority of contacts
• 6% escalation rate to live agents, indicating high chatbot confidence and containment effectiveness
• 24/7 availability improved customer satisfaction by removing wait-time friction during evenings, weekends, and across time zones

The consistently low escalation rate reflects both the quality of the NLP model and the depth of system integration enabling accurate, policy-relevant responses.

• Having a defined technology direction is not the same as having implementation capability — P&C firms should expect to source linguistics, NLP training, and UX expertise externally.
• A DevOps delivery model allows the chatbot to improve incrementally based on real conversation data rather than requiring a perfect solution at launch.
• Containment rate (inquiries resolved without escalation) is the primary operational metric to track — a 50%+ rate at this volume directly drives cost savings.
• Structuring the vendor engagement as a trusted advisor relationship enables the solution to evolve as LLM capabilities advance, protecting the initial investment.
• 24/7 availability is a differentiator in P&C where customers frequently need support outside business hours following incidents.