NatWest Group boosts mortgage customer loyalty 20% with AI-powered call center assistant 'Marge'

For mortgage lenders like NatWest — serving approximately 19 million customers across the UK and Ireland — the home buying process demands that call center agents apply precise, up-to-date policy knowledge to each customer's unique situation in real time. Mortgage regulations, product terms, and internal processes change continuously, making it impossible to keep agents current through periodic training alone. Without a single authoritative source of guidance, agents frequently encountered knowledge gaps mid-call, reducing their confidence and extending handle times. For customers navigating one of the most consequential financial decisions of their lives, uncertainty from the agent translated directly into eroded trust and lower satisfaction scores.

NatWest partnered with IBM Consulting to co-create Marge, an AI-powered digital mortgage support platform built on IBM watsonx Assistant running on IBM Cloud. Rather than replacing agents, Marge acts as a real-time knowledge layer: during live customer calls, agents type keywords into a console and receive immediate, accurate policy guidance. Marge was built directly on the cloud and embedded within NatWest's existing data infrastructure, giving it access to content updates every minute as regulations and products evolve. To drive adoption, NatWest intentionally personified the assistant as a named team member with an evolving personality — positioning Marge as a colleague rather than a tool. The platform continuously learns from new data and customer interactions, reducing reliance on manual knowledge-base maintenance.

Following deployment, NatWest recorded measurable improvements across both customer and operational metrics:

• 20% improvement in customer Net Promoter Score (NPS) — a key indicator of loyalty and likelihood to recommend
• 10% reduction in average call duration — driven by faster, more confident agent responses

Beyond the headline numbers, agents reported a significantly higher level of confidence when advising customers on mortgage decisions. According to MaryAnn Fleming, Head of Homebuying Services at NatWest, that agent confidence translated directly into greater customer trust throughout the home buying journey. The platform's continuous learning capability means accuracy improves over time without manual retraining cycles.

• Agent-assist AI outperforms self-service deflection in high-stakes, complex transactions like mortgages — confidence at the agent level drives customer loyalty more than automation alone.
• Personifying an AI tool with a name and identity accelerates internal adoption by framing it as a team member rather than a system to learn.
• Real-time data ingestion is essential in regulated industries; a knowledge base that lags behind regulatory changes becomes a liability, not an asset.
• Embedding AI within existing data structures reduces deployment friction and ensures the tool reflects the organization's actual policies from day one.