AXA

In Property & Casualty insurance, average handle time and first-contact resolution are direct drivers of operating cost and customer retention — particularly at policy renewal. AXA's UK retail business unit, covering car, home, health, and business insurance for millions of customers, faced compounding fragmentation after consolidating three divisions into one operation. Voice recording analysis and desktop analytics ran on separate, unintegrated tools, requiring costly data bridges and undermining confidence in reporting. Manual quality review sampled only a small fraction of calls, leaving compliance gaps and process inefficiencies undetected. A concurrent CRM rollout intensified the pressure, unexpectedly increasing average handle time by 25% across the consolidated operation.

AXA migrated its contact center analytics to the Verint Open Platform in the cloud, consolidating fragmented tooling into a unified stack. Verint Da Vinci AI-powered Speech Analytics and Verint Desktop and Process Analytics were integrated to deliver a continuous, cross-channel view of each customer interaction. Rather than importing existing call categories, AXA rebuilt them as smaller, process-aligned modules — a deliberate redesign that lifted NLP transcription accuracy to 95% and increased confidence in AI-generated insights across the business. Verint Performance and Compliance Scoring Bots then replaced manual call sampling with automated, AI-driven quality evaluation at scale. For First Notice of Loss (FNOL) claims — the most variable and operationally complex call type in P&C — a complexity matrix was introduced to model predicted handle time and support more precise workforce planning and resource allocation.

The most immediate result was a 23% reduction in customer renewal average handle time — equivalent to 182 seconds per call — achieved within two months of deployment. Redirecting suppliers to a self-service portal drove a 34% decline in inbound supplier call volume, reclaiming significant agent capacity for customer-facing interactions. First contact resolution climbed from 87% to 93%, reflecting improved agent guidance and more accurate call routing. Customer satisfaction scores followed across the claims operation:

• Home Claims NPS: improved by 97%
• Motor Claims NPS: increased by 26%
• Transcription accuracy: reached 95% after call category redesign

• Unifying voice and desktop analytics on a single platform eliminates integration overhead and restores confidence in operational data — a prerequisite for reliable AI-driven action in P&C contact centers.
• Rebuilding NLP call categories as smaller, process-aligned modules rather than broad topics is directly responsible for transcription accuracy gains; AXA's 95% accuracy required this architectural decision.
• Automated AI quality scoring closes the compliance coverage gap left by manual sampling, without requiring additional headcount.
• For high-complexity claim types like FNOL, a predictive handle-time model improves workforce planning accuracy beyond what aggregate AHT targets alone can achieve.
• Self-service deflection for third-party contacts such as suppliers can yield outsized volume reductions, freeing live agent capacity for higher-value customer interactions.