Commonwealth Bank moves 84.6% of customer queries to self-service: how an AI agent transformed retail support

At Commonwealth Bank, one of Australia's largest financial holdings, customer support in retail banking has always been a complex challenge. Tens of thousands of requests daily, each requiring attention and quick resolution. Following the implementation of an AI agent that took over initial inquiry processing, 84.6% of requests are now resolved through self-service in digital channels, and the speed of problem resolution has significantly increased.

In a large bank, customer support is a funnel into which thousands of diverse questions flow. Each requires navigation through complex internal systems, switching between departments, and manual information gathering. This is expensive, slow, and often leads to customer frustration. But crucially, it's a funnel that can be automated, changing the approach to service.

The pain points of retail banking customer support

Retail banking involves millions of customers and thousands of touchpoints. Questions can vary widely: from balance inquiries and password changes to disputing transactions and applying for loans. Traditional support requires a large staff of operators who often spend time on routine tasks: customer identification, searching for information in various systems, and redirecting inquiries. This leads to long wait times, the need to repeatedly explain problems to different employees, and consequently, a decrease in customer satisfaction.

The problem is exacerbated when requests come through digital channels, where customers expect instant replies. In such conditions, every unresolved request is not just a waste of time, but a risk of losing loyalty.

Why classic chatbots fell short

Commonwealth Bank, like many large financial organizations, already used chatbots and automation systems. However, they were often limited by rigid scripts and could not effectively handle complex or non-standard requests. As soon as a question went beyond predefined branches, the bot became useless, and the request still went to a human. This created a "bottleneck" effect and did not solve the problem of operator overload.

The bank needed not just a bot, but an intelligent agent capable of understanding customer intent, query context, and dynamically directing it to the most appropriate solution, whether another AI system or a live specialist. This led to the idea of creating a central AI orchestration agent.

How the AI orchestration agent was designed

The AI agent was conceived as a central hub that interprets customer intent and dynamically routes the request. Its main task is to understand exactly what the customer needs and direct them to the optimal solution. This could be:

• Conversational AI. For simple and frequently asked questions, where information is publicly available or in knowledge bases.
• Human specialist. For complex, sensitive, or non-standard situations requiring empathy and deep contextual understanding. In this case, the AI agent remains in the background, providing the specialist with a conversation summary and suggesting possible responses.
• Another AI system. For example, for automated transaction execution or personalized information provision.

The key was intelligent, context-driven routing, which allows the system to continuously learn and improve service quality.

Implementation and scaling

The development of the AI agent took two years and began with a focus on retail banking, as the most widespread and critical area. After a successful pilot launch and demonstration of effectiveness, the bank began to scale the solution. Today, the system has performed so well that Commonwealth Bank plans to expand its application to other areas of its operations, beyond retail banking. This includes support for voice bots, multi-agent workflows, and ultimately, the creation of a bank-wide conversational platform.

Results

General Manager of Assisted Customer Experiences Rachel Round noted that the system provided "a step-change in how effectively customer enquiries are being resolved through [the bank’s] digital and messaging channels." 84.6% of self-service interactions in messengers represents not only massive resource savings but also a significant improvement in customer experience, as clients receive fast and accurate answers without waiting for an operator.

How to implement this in your company

The Commonwealth Bank case shows that AI agents can radically change the approach to customer support in large organizations. If your company faces similar challenges with processing customer inquiries, here's where you can start:

• Identify "hot spots." Pinpoint the most frequent and routine inquiries that consume a lot of operator time and can be automated.
• Start with an orchestrator. Instead of creating many separate bots, focus on an AI agent that will route inquiries to the right resources—whether another AI or a human.
• Integrate with existing systems. For effective operation, the AI agent needs access to knowledge bases, CRM, and other internal systems to gather context and provide accurate answers.
• Train and scale gradually. Start with one area, gather feedback, train the agent on real data, and only then scale to other business areas.

If this case sounds like what's happening in your company, our manager can help: he'll analyze your business and niche for free and point out where an AI agent would bring a real result in your case. Message the manager