Royal Bank of Scotland (RBS)

Royal Bank of Scotland faced a service capacity crisis driven by its own digital growth. With over 18 million customers and a mobile app base that had expanded to 6.1 million active users, demand for real-time support was outpacing what phone-based channels could absorb. Traditional voice support is inherently single-threaded — one agent, one customer, one call — making it structurally unsuited to the volume and pace of digital banking interactions. As customers increasingly expected instant responses through mobile and web channels, RBS needed a fundamentally different service model or risk degraded customer experience and unsustainable agent headcount growth.

RBS deployed LivePerson Conversational Cloud as the foundation of its digital service transformation, integrating it across web chat, in-app messaging, and proactive outreach channels. At the core was Cora, a virtual assistant built on IBM Watson, designed to handle common banking queries autonomously before escalating to human agents when needed. The platform enabled smooth bot-to-agent handoff, preserving conversation context so customers never had to repeat themselves. By deploying conversational AI natively within the RBS mobile app alongside web channels, the bank could meet customers in the digital environments they already used. Proactive messaging capabilities allowed RBS to reach customers with relevant updates without waiting for inbound contacts, further shifting volume away from reactive voice queues.

The impact on service scale and efficiency was measurable and significant. Cora and the automation layer now resolve 27% of all conversations without agent involvement, handling over 40,000 customers daily through automated responses. Combined messaging and automation volumes exceed 63,000 conversations per week, a throughput level the bank's voice infrastructure could not have matched at comparable cost. The headline operational gain: messaging-based service is 1.8x more efficient than voice, reflecting the concurrency advantages of asynchronous digital channels. Qualitatively, agents were freed from high-volume routine queries and redirected toward complex, relationship-sensitive interactions — a structural improvement in how human capacity is allocated.

• Automation and human agents are most effective as a layered system: route routine queries to bots, reserve agents for complexity, and ensure handoff preserves full conversation context.
• Mobile app integration is a multiplier — deploying conversational AI inside the channel customers already use daily removes friction and drives adoption without requiring behavior change.
• Proactive outreach shifts service from reactive to anticipatory, reducing inbound load before it builds.
• The 1.8x voice efficiency gain reflects concurrency: one messaging agent can manage multiple simultaneous conversations in ways a voice agent cannot — model staffing assumptions accordingly.
• AI virtual assistants require a clear escalation path to human agents; containment rates improve only when customers trust that help is available if the bot reaches its limits.