Key takeaways from our Bud x PA Report on GenAI in banking

Earlier this year, we co-authored a report ("How to use GenAI to multiply customer insights from transaction data") with PA Consulting, with contributions from Google Cloud, DataStax and Zup Innovation. Exploring the role of GenAI in transforming the retail banking sector, we shared our insights alongside industry experts into laying the foundations for success in GenAI banking. 

Here’s a summary of the full report.

GenAI and its potential for the 'AI-first' bank

The potential for Generative AI (GenAI) to transform banking is huge and we’re starting to see the emergence of the ‘intelligent bank,’ which adopts an ‘AI-first’ mindset in the same way that banks have adopted a ‘digital-first’ or ‘mobile-first’ mindset over the last 10 years.

GenAI is already being used in many of banks’ strategic and operational decision-making processes to generate deeper insight for both humans and agentive models. This enables step changes in business agility, hyper-personalized customer service and scaled up automation.

The two largest opportunity spaces for banks are to:

• create semi-autonomous middle-office processes, and
• build more sophisticated customer engagement and assistance systems.

As the era of the intelligent enterprise gathers pace, banks that are effective at adopting an ‘AI-first’ mindset at the enterprise scale will succeed, while others risk losing significant market share and potentially failing altogether.

Transaction enrichment as the key to GenAI success

The core focus of GenAI conversations in the banking context is on large language models (LLMs) which are great at dealing with text information, but most effective when working with natural language. This poses a challenge for banks because a lot of data needs to be processed to be useful for GenAI. 

For transactions, this means adding dimensions that can be described with natural language, and with as much granularity as possible, to ensure that all potential patterns and matches will be found. If trained on general financial knowledge and perhaps internal procedures, a GenAI model will be far more effective as the processed data would be explained using language that matches the type it already knows. 

The easiest way to ensure this outcome is to feed the model with transactions that have categories – meaning easy to understand words that can be supplemented with short definitions for even more context.

But why do it?

Because transaction data enrichment done well allows banks to build a very accurate picture of the customer by translating account statement line items into customer insights, such as spending commitments (for example as a parent or pet owner), behavioral patterns, attitudes, priorities, income stability and total wealth. These insights can be used to deeply personalize customer service, marketing and risk scoring.

Seizing the competitive edge with GenAI

The banking industry stands at a pivotal moment. GenAI isn’t just a future possibility; it’s a present reality that’s reshaping the financial landscape. Early adopters are already harnessing GenAI to transform customer experiences, streamline operations and unlock unprecedented levels of insight from their data.

Early adoption as a competitive advantage

The banks that embrace GenAI now will set the pace for the industry. They’ll gain a competitive edge by delivering hyper-personalized services, proactively identifying risks and tailoring products to individual needs with unparalleled precision.

Data as the differentiator

The quality and depth of our customer data are the foundations for GenAI success. Banks that have invested in enriching their transaction data (through robust merchant identification, categorization and contextualization) are primed to extract maximum value from GenAI applications. Delaying this enrichment means delaying the transformative potential of GenAI.

Shifting customer expectations

Today’s customers expect intuitive, personalized experiences. GenAI enables banks to meet and exceed these expectations, fostering loyalty and attracting new customers. Banks that lag behind risk alienating their client base. 

Operational efficiency

GenAI-powered automation can streamline processes, reduce manual intervention and free up employees to focus on higher-value tasks. This translates to cost savings and improved operational efficiency.

Key challenges to implementing GenAI in banking

1. Data privacy and security: Because banks handle highly sensitive customer data, it’s essential for GenAI systems to adhere to strict data privacy regulations like GDPR. Protecting against data breaches, unauthorized access and adversarial or model inversion attacks is critical as these threats can compromise AI model integrity, expose confidential information and disrupt operations.
2. Bias and fairness: GenAI systems must be trained on diverse datasets to avoid biases that could lead to unfair treatment of certain customer groups. Ongoing monitoring is essential to detect and mitigate bias, maintain customer trust and prevent discrimination.
3. Data quality and integration challenges: Effective GenAI depends on high-quality, enriched data. Banks frequently encounter hurdles when integrating diverse data sources and ensuring data consistency, which directly affects the performance and reliability of AI models
4. Data drift: The underlying data distributions may change over time, leading to model degradation. Regular monitoring and retraining of models are necessary to address data drift. This helps adapt to changing patterns and improve accuracy.
5. Scalability and performance: Scaling GenAI solutions to handle the massive volume of transactional data in real time requires robust and scalable infrastructure. Performance bottlenecks can hinder the effectiveness of these systems. Shifting from LAB to production introduces challenges such as maintaining model accuracy and performance under real-world conditions.
6. Change management: Introducing GenAI systems involves significant changes in workflows and processes. Effective change management strategies and training are necessary to ensure smooth adoption and to address resistance from employees
7. Interoperability: Ensuring that GenAI solutions seamlessly integrate with existing banking systems, data sets and workflows can be complex. Compatibility issues can slow down the deployment process.

Taking the next step in your GenAI journey

Dive into the full report and discover actionable insights around:

• Scaling GenAI safely to multiply the business value from transaction data
• Data architecture for GenAI adoption
• Creating a transaction data and insights Centre of Excellence, and more.

Full report download: How to use GenAI to multiply customer insights from transaction data