McKinsey and Capital One: Customer AI, a Driver of Innovation

The Limited Impact of Digital Investments According to McKinsey

According to a study conducted by McKinsey, despite increasing digitization, companies are only able to capture a fraction of the expected benefits from their digital investments. In fact, they capture less than one-third of the anticipated value from these investments. This observation stems from the fact that many prioritize the development of their technological capabilities before considering practical applications, rather than starting from the actual needs of customers to develop tailored solutions. This approach can lead to incoherent solutions, fragmented customer experiences, and ultimately, digital transformations that fail to achieve their objectives.

A Customer-Centric Approach for Tangible Results

Conversely, some organizations achieve significant results by reversing this logic. They adopt a customer-centric engineering strategy, where customers become the starting point for any technological transformation. This method involves designing products and services with the expectations and challenges of customers in mind from the outset. Development teams then work in an agile manner to determine the necessary steps to create solutions that meet these expectations.

Ashish Agrawal, Senior Vice President at Capital One, emphasizes that bringing engineers closer to customers fosters lateral innovation. This allows engineers to approach problems from a new perspective, often one that is novel from a sales or product viewpoint, thereby creating a multiplier effect in terms of innovation.

The Crucial Importance of Customer Orientation

Agrawal explains that engineers, as problem solvers by nature, are particularly effective when they understand the challenges customers face. By being close to systems and data, they can design solutions that directly address customer needs. This proximity to customers has a motivating effect on engineers, who see the direct impact of their contributions on users' lives.

To encourage this culture, Capital One has established clear objectives for its engineers, urging them to establish multiple touchpoints with customers each year. These interactions take various forms, such as:

• Digital empathy sessions to observe user journeys and identify pain points.
• Integrated customer support to better understand service needs.
• Engineering shadowing, where engineers participate in calls or site visits with sales and support staff.
• Hackathon competitions to develop solutions to real problems faced by customers.

Opportunities Offered by Customer-Centric AI

According to Agrawal, one of the biggest challenges for engineers in large companies is the lack of direct access to customers, which complicates problem identification and solution innovation. However, AI has accelerated both challenges and opportunities. The product launch cycle has become much faster, and engineers, having privileged access to the data fueling AI, can apply data techniques more quickly to solve customer problems.

A concrete example is customer service, where AI can instantly summarize conversations and provide agents with precise context on customer requests and actions to take. Agentic AI can also ask relevant follow-up questions, which would otherwise require considerable time from human agents.

Agrawal points out that in a data-rich ecosystem with high-quality data, agentic tools enable a shift from incremental fixes to rapid, large-scale transformation. By investing in data and AI tools, and focusing on rapid experimentation, the deployment cycle for solutions can be significantly accelerated.

For instance, Capital One has used customer insights to develop an innovative multi-agent AI framework called Chat Concierge. This tool enhances the customer experience for car buyers and dealers by allowing, in a single conversation, the comparison of vehicles, assistance in decision-making, and scheduling test drives or appointments with sellers.

Towards an AI-Centric Mindset

A recent survey by MIT Technology Review Insights reveals that 70% of executives claim their company uses agentic AI to some extent. Nearly half of them believe these systems improve fraud detection, security, reduce costs, increase efficiency, and enhance customer experience.

Looking ahead, these results seem promising. More than half of banking executives plan to further improve fraud detection, security, and customer experience. Use cases for agentic AI that show strong potential include responding to customer service requests, adjusting bill payments, or extracting key terms from financial agreements.

To achieve these goals, companies must adopt an AI-centric mindset, shifting from enhancing an existing product to fundamentally reinventing the problems and needs of users through AI capabilities.

Agrawal recommends several best practices:

• Reinvent the core function of AI to solve user problems. The true value lies in addressing significant issues for customers, ensuring that innovation is not only rapid but transformative.

• Start with high-quality, well-governed data. Rigorous data preparation is essential to orchestrate the agentic loop, enabling the perception, reasoning, and execution necessary to solve customer problems.

• Rebuild workflows with integrated AI from the outset. Agentic systems require rigorous oversight. A well-governed data ecosystem and responsible AI standards are crucial for establishing trust.

• Form a cross-functional team involving data science, engineering, product, design, and other partners. It is important to adopt a gradual approach if one is new to AI, rather than diving in headfirst.

Ultimately, successful transformation relies on empowering engineers and partner teams to develop technological solutions based on customer needs, rather than starting from technological capabilities to find applications. By adopting a customer-centric approach, organizations can reinvent the customer experience from within and place the customer at the center of their strategy from the very beginning.