How is generative AI transforming retail banking?

With its ability to quickly parse vast amounts of data, generative AI is a powerful asset to data-rich, data-reliant sectors like retail banking. The technology enhances customer engagement, anticipating users’ needs with hyperpersonalized solutions. What’s more, generative AI automates low-value tasks, freeing up employees to focus on more complex activities with higher impact, and unlocks significant productivity and efficiency gains.

How does generative AI work across banking functions?

Generative AI uses machine learning-based large language models, transaction analysis, and natural language interfaces to generate new content such as responses and insights. Typical use cases within the sector leverage:

• Summarization—generating concise and precise summaries of a collection of text or other media (e.g., insights from customer communications)
• Generation—creating new digital images, video, audio, or text (e.g., personalized ads)
• Interaction—communicating between a user and a computer system (e.g., customer service chat)
• Knowledge management—surfacing information and insights embedded in all manner of data (e.g., documents, e-mails, transcripts)

Deploying generative AI in banking requires navigating a complex landscape of technical, ethical, regulatory, and security considerations. Among the key components:

• Enterprise API gateway serves as the secure front door between client applications and backend services, streamlining API development and management.
• Data ingestion centralizes diverse data sources for streamlined processing and analysis.
• Data serving delivers data efficiently to applications, services, or end users when and where it’s needed.
• Generative AI services integrate pretrained or custom-built models to power specific use cases across the organization.

What are the core generative AI use cases in retail banking?

Retail banks are harnessing intelligent chatbots, voicebots, and agent tools to improve and accelerate everything from brand awareness to customer onboarding. Generative AI is already reshaping the sector through several key use case categories.

• Marketing: Generate hyperpersonalized marketing assets such as emails, tailored product descriptions, and customized landing pages.
• Onboarding: Automate client onboarding and fraud detection with advanced document verification, intelligent document processing (IDP), and biometric checks; develop banking product Q&A for new customers, personalized customer activation, and targeted offers.
• Call center activities: Summarize customer interactions based on call transcripts; conduct sentiment analysis with predictive NPS® score; identify root cause of customer dissatisfaction and suggest next steps.
• Document and knowledge management: Generate synthesized key points from structured input (e.g., internal documentation) or unstructured input (e.g., emails from employees) and respond to specific requests and questions related to policies, products, and services.
• Risk and compliance: Detect early or missed payments, monitor high-risk profiles, and ensure compliance with real-time data analysis and reconciliation.

How can generative AI drive value in retail banking?

Banks with a strong focus on technology consistently outperform the competition. It’s no surprise that industry leaders are embracing AI as a path to efficiency, improved customer service, and revenue growth.

Our analysis of generative AI in financial services shows significant potential reduction to the cost base in two to three years, including 20% to 30% in customer service and 15% to 25% in risk and compliance. Broadly, today’s generative AI winners are seeing:

• Top-line growth. AI enables greater focus on growth-oriented tasks from the front line and creates opportunities for business growth through personalization.
• Transformation cost reduction. AI impacts the cost base by automating repeatable, low-value tasks and increasing customer use of digital self-service channels.
• Enhanced customer satisfaction. AI improves customer satisfaction through more compelling experiences, quicker service times, and better issue resolution.

What risks and governance challenges exist for banks adopting gen AI?

Generative AI adoption comes with real hurdles. Training and running large language models can be time- and resource-intensive. And as new capabilities emerge, companies often need to upskill teams or hire new talent to manage them effectively. Smart retail banks will segment their risks and develop mitigation tactics for each category. Common risks include the following:

• Accuracy. Baseline models can present inaccurate or false output.
• Quality issues. Language/image output does not always meet requirements.
• Rights. Ownership of output remains somewhat ambiguous.
• Risk and compliance. AI could be biased and respond inappropriately to users.
• Security. Integration with third-party AI capabilities could introduce model theft or misuse and cause privacy breaches.
• Data privacy. Use of customer data to train and improve models could expose companies to unauthorized data usage, reidentification risks, and data leakage.
• Deployment model. Dependency on third-party providers may cause latency and service disruption.

While these challenges are real, they are addressable. Our advice to retail banks?

• Establish guardrails around information sources and input design.
• Pre-process data to reduce inaccuracies; instruct models to identify the source of data alongside answers to verify authenticity.
• Retain human oversight within training and inferencing (i.e., human in the loop) to ensure response validity.

What are the future trends in generative AI for retail banking?

Across retail banking and beyond, generative AI is moving from a cost-cutting tool to a catalyst for better customer experiences. For example, Capital One launched Chat Concierge, an AI agent that aims to relieve the “cognitive burden” of purchasing a vehicle by managing tasks from estimating the value of a trade-in to scheduling appointments with sales staff.

Also gaining momentum? Voice activation, greater task and workflow automation, more robust fraud detection, and the ability to pinpoint market gaps and customer needs to develop new products and services.

How can banks scale generative AI safely and strategically?

To surface the most high-value generative AI use cases, consider three factors:

• Value-at-stake—accounting for both the initial implementation and scalability
• Ease of building, considering the complexity of the interaction and the integrations required to support it
• The inherent risk given the nature of the interaction and the data being used (confidential, PII, etc.)

From this starting point, align on initial use cases that balance these considerations while adhering to responsible AI principles. In parallel, build out a longer list of use cases for a strategic roadmap. This charter should also include governance, decision rights, program design, operating model design, and change management support. Finally, ensure your organization has the necessary capabilities to scale for long-term success.

Build the bank of the future with generative AI

When it comes to building the bank of the future, early movers will have the advantage. Consider Bradesco, which pioneered the use of AI in the financial sector nearly a decade ago. Today, the Latin American bank boasts a generative AI chatbot that resolves customer problems without human intervention in 90% of cases, serving millions of customers every day.

Capturing value from generative AI requires speed—and speed hinges on making the right organizational choices. By asking the right questions, running real experiments, and scaling early wins, you lay the groundwork for the bank of the future.

Seeking additional context? Explore our AI services along with published insights and case studies on generative AI in financial services and banking, including trends and data drawn from real-world deployments.