DNB

As Norway's largest bank, DNB manages customer service interactions at significant scale across its retail banking operations. The bank's contact center was fielding a high volume of routine, low-complexity inquiries — account balances, payment questions, card services — that consumed the time of skilled human agents and created bottlenecks during peak periods. This imbalance between inquiry type and agent capability constrained DNB's ability to deliver consistent response times while limiting human agents' capacity to handle higher-value, more nuanced customer needs. Without automation, scaling service quality to match growing digital channel demand would require proportional headcount increases.

DNB partnered with Boost.ai to deploy a conversational AI virtual agent named Aino as the primary first point of contact across its digital chat channels. Built on Boost.ai's natural language understanding platform, Aino was integrated into DNB's existing customer service infrastructure to handle routine inquiries autonomously and route complex cases to human agents with full context. In parallel, DNB deployed a second Boost.ai-powered agent for internal use, enabling employees to access information and resolve queries more efficiently. Recognizing that ongoing model quality depends on operational discipline, DNB made a deliberate investment in staffing 15 dedicated AI trainers responsible for continuously refining Aino's responses, expanding its knowledge base, and managing escalation logic.

Aino now autonomously resolves more than 50% of all inbound chat traffic and handles over 20% of total customer service requests — having served more than one million customers since deployment. The employee-facing virtual agent delivered an 80% improvement in accuracy rates, enabling staff to retrieve information and respond to customers more reliably. Qualitative outcomes included measurable gains in customer satisfaction scores and a meaningful shift in how human agents spend their time, with routine inquiry volume absorbed by automation freeing agents for relationship-intensive and complex cases. The dual deployment — customer-facing and employee-facing — created compounding efficiency gains across the service operation.

• Staff AI trainers as a dedicated function: DNB's investment in 15 full-time AI trainers signals that virtual agent quality is an ongoing operational commitment, not a post-launch afterthought.
• Deploy on both sides of the conversation: Pairing a customer-facing agent with an employee-facing counterpart multiplies efficiency gains — agents become faster and more accurate while customers receive faster resolutions.
• Measure success across two dimensions: Track customer-facing metrics (automation rate, satisfaction scores) alongside employee-facing metrics (accuracy, handle time) to capture the full impact.
• Volume and complexity routing is the core design problem: Success depends on correctly distinguishing automatable inquiries from cases requiring human judgment — get this wrong and automation erodes satisfaction rather than improving it.