National Bank of Canada scales agentic AI from advisers to customers

As customer expectations evolved, NBC saw an opportunity to help branch advisers access trusted home-finance information faster and more consistently. Advisers were already drawing on internal knowledge repositories with a manual search to answer customer questions about home finance.  The Bank aimed to make that process even more seamless. By improving access to answers, NBC could reduce escalations to the internal support center, which was contributing to call volume growth, and give advisers more time for meaningful customer conversations.

To achieve that goal, NBC invested in advanced search capabilities. Search improved access and gave advisers faster answers. The Bank then deployed an intent-based chatbot and, building on the results in call volume and satisfaction, explored the agentic capabilities of Amazon Bedrock, a service for building generative AI applications and agents.

Building a compliant, robust agentic AI solution took proven experience and strategic support. To create an AI agent for branch advisers, NBC’s AI Factory team worked alongside AWS specialists, who provided hands-on guidance and support for AI and machine learning services. The team built the solution by using Anthropic’s Claude on Amazon Bedrock. The team also relied on Amazon OpenSearch Service to power knowledge access with a near real-time retrieval engine for AI, search, and analytics.

To accelerate the cloud journey with an outcome-focused transformation methodology, NBC’s engineers joined AWS experts in the AWS Experience-Based Acceleration (AWS EBA) program. Through the program, the Bank gained the framework and hands-on AWS expertise to rearchitect its single-purpose agent into a coordinated multiagent system. This includes domain-specialized agents that draw on separate knowledge bases to enhance NBC’s AI-powered contact center capabilities. The system spans banking operations, personal financing, residential financing, credit card services, investment advisory, and application support. The specialized agents collaborate to provide comprehensive and accurate responses across all customer questions that advisers receive.

The Bank made a strategic choice to start internally, proving the technology with employees first and building organizational confidence before considering public deployments. NBC began with a pilot that involved 150 branch advisers to gather feedback and optimize performance. The agent responded to requests in both French and English and helped advisers retrieve information with minimal errors and focus on complex customer needs.

NBC scaled the deployment to 700 advisers 1 month after the pilot. The agentic solution surpassed both the previous solution’s performance and NBC’s accuracy target across all banking segments. Within 3 months, the Bank saw a 10–15 percent net reduction in call-center volume—even as mortgage and business volume surged by about 40–50 percent. The solution reached a 97 percent satisfaction rate, and 75 percent of advisers reported daily time savings. “We wanted a call center that was focused on outbound calls or complex situations,” says Julien Crowe, Senior Director AI at NBC. “But that we got there within a few months was a great surprise for us.”

To extend AI capabilities to customers, NBC launched a conversational AI agent on its website for small, medium, and large businesses. For commercial call agents, this translated into a volume reduction of 25 percent. Now, commercial clients have self-service access to information, and branch clients have more fluid interactions. “Branch advisers can now work directly with clients,” says Julien Crowe. “Built on AWS, our agentic AI solution gives advisers more freedom to guide conversations effectively.”

There was also a net reduction of 43 percent in calls to the internal support center, even while communication from all sources and channels increased by over 50 percent. AI handled the surge, absorbing routine inquiries and reducing reliance on phone-based escalation, limiting a proportional spike in support-center calls.

In 2025, the bank retired the dedicated inbound call-center line used by branch advisers for home-finance support. Its agentic system now serves as the first point of contact in NBC’s contact center, helping advisers find answers faster and focus more on customer conversations. The solution has also expanded to other business areas, including a human resources agent that responds to employees’ requests with 85 percent accuracy and a 4.3 out of 5 satisfaction score.

For its agentic solution, NBC won the Best Gen-AI Chatbot Initiative award at The Digital Banker’s Global Retail Banking Innovation Awards 2025. NBC is extending its agentic approach toward workflow automation, experimenting with transactional agents for back-office use cases. The Bank is also bringing AI into its software development lifecycle, where agents support analysis, prototyping, and code comprehension, cutting delivery time from weeks to days.