Aviva

Aviva, the UK's largest general insurer, faced claims costs rising at 11% above CPI by 2023 — up from 6% above CPI in 2019. The core challenge was structural: settling a P&C claim involves dozens of interdependent decisions across liability assessment, fraud detection, repair routing, and final settlement. Each decision point carried downstream cost and customer experience consequences. Yet despite this complexity, the insurance industry had been slow to apply AI to claims, focusing instead on pricing and retention. For Aviva, claims represent the highest-stakes customer interaction — often involving distress, bodily harm, or significant financial loss — making the cost of slow, inaccurate decisions especially high.

Aviva partnered with McKinsey's QuantumBlack (AI) and Orphoz (implementation) to execute a full-domain transformation rather than targeting isolated pain points. A joint team of 50+ data scientists, engineers, business leaders, and change professionals built and deployed 80+ predictive ML models spanning the entire claims function — covering liability assessment, fraud signals, repair routing, settlement valuation, and customer journey orchestration. A 'double helix' architecture was embedded to allow seamless switching between automated digital workflows and human intervention based on case complexity — personal injury claims, for instance, default to human handlers. Translators bridged technologists and frontline claims staff throughout development to ensure each model reflected real operational needs. Over 40,000 hours of training were delivered to embed a data-as-asset mindset and agile ways of working across the organization.

The transformation delivered measurable improvements across customer experience, operational efficiency, and employee outcomes:

• 23 days reduction in average liability assessment time for complex cases
• 30% improvement in repair routing accuracy
• 65% reduction in customer complaints
• 7x improvement in Net Promoter Score
• Employee engagement scores more than doubled, reaching an all-time high
• Use of recycled parts in vehicle repairs tripled, reducing costs and environmental impact

The routing accuracy gains translated directly into fewer escalations and faster resolutions. Employee outcomes improved in parallel with customer metrics — a result Aviva attributed to giving claims professionals better tools, not replacing their judgment.

• Domain-wide deployment outperforms point solutions: Targeting isolated steps in claims produces marginal gains; simultaneous AI coverage across liability, fraud, routing, and settlement compounds the impact.
• Cultural transformation is load-bearing: 40,000+ hours of training and agile operating model changes were prerequisites — not afterthoughts — to sustaining AI adoption.
• The 'double helix' model resolves the automation tradeoff: Designing explicit handoff points between digital workflows and human judgment avoids the false choice between efficiency and quality.
• Translator roles accelerate model accuracy: Embedding change professionals as bridges between data scientists and frontline staff reduced rework and improved model fit to real operational decisions.
• Measure employee outcomes alongside customer metrics: Engagement gains signal whether AI is augmenting or degrading frontline work — a leading indicator of sustainable adoption.