The True Potential of LLMs and AI in Banking | Lloyds, ING, and BBD | FF Virtual Arena #334

Do we need to take a fresh look at AI in banking?

In this fascinating FF Virtual Arena, our esteemed panel of guests investigate how AI and Large Language Models could really impact banking and financial services.

Going beyond the hype, Ranil Boteju from Lloyds, Kevin Staples from BBD and Marco Li Mandri from ING, unpack the practical ways it can impact digital transformation today, the potential for radical changes in customer experience and the barriers to adoption.

There’s also some exciting predictions for the next 5 years.

Is AI Adoption Coming into Land?

To kick off the conversation Ranil Boteju reminded us that AI isn’t new to banks. AI and machine learning (ML) have been integral to banking systems for decades, especially in processes like Know Your Customer (KYC), identity verification, and automation. However, the rise of generative AI, popularized by tools like ChatGPT, has brought new opportunities and challenges. Unlike previously hyped technologies such as blockchain and quantum computing, generative AI has seen swift, organic adoption across various departments within his bank, Lloyds. 

Engineers and teams independently explored AI’s potential, resulting in a surprising number of practical use cases in areas like complaints management, underwriting, and claims. But there’s caution here, particularly around exposing AI directly to customers, citing the need for robust guardrails and governance structures to mitigate risks like misinformation and bias.

AI in Contact Centers and the Rise of The Digital Twin

Marco Li Mandri provides some more use cases, saying that recent advancements in generative AI are transforming areas like contact centers and marketing personalization. AI could support analysts, streamline processes in software engineering, and improve sustainability efforts in wholesale banking. One of the most promising applications, according to Li Mandri, is AI’s ability to accelerate data extraction in lending, making processes more efficient and aligning with the bank’s sustainability goals.

Really at the heart of all this are LLMs or Large Language Models which are powering a lot of this advancement. 

Kevin Staples from technology consultancy BBD discussed the broader adoption of LLMs in banking, classifying AI’s use in three categories: customer-facing applications, operational efficiency, and back-office functions. For Staples, the most exciting prospect lies in customer-facing innovations, specifically the concept of a “digital twin” for personal banking. He envisions a future where every customer has a digital counterpart that can provide personalized financial advice, essentially replicating the human touch of a personal banker.

While Staples acknowledged that we aren’t there yet, he stressed that LLMs could revolutionize customer interactions by enabling banks to offer highly personalized, automated services. However, one of the complexities in achieving this is the level of sophistication needed for LLMs to understand and cater to individual customer needs without compromising security or regulatory compliance.

Governance and Risk Mitigation Is Key

There are also risks to the adoption of AI and LLMs in banking and Boteju discusses Lloyds’ approach to risk management, particularly around generative AI, which introduces new concerns like hallucination, misinformation, and security vulnerabilities. Frameworks like the U.S. National Institute of Standards and Technology (NIST) are guiding their strategies. They employ a combination of manual and automated controls to mitigate these risks, including “human-in-the-loop” systems where human oversight ensures that AI outputs are accurate and reliable before they are deployed at scale.

Staples also highlights the importance of having a robust governance framework in place to monitor AI systems, especially as LLMs can sometimes generate unpredictable or inaccurate results. He spoke of the challenges banks face in integrating AI with their existing infrastructure, stressing that a strong collaboration between data scientists, software engineers, and risk managers is crucial to ensuring that AI is deployed responsibly.

Human Roles in the AI Era

One of the big concerns around the growth of AI of course, is the impact it will have on human involvement and jobs. But despite AI’s growing prevalence, the panelists were unanimous in their belief that human expertise will remain essential. Li Mandri stressed the need for educating employees across all levels about the risks and opportunities AI brings, especially as banks move toward implementing AI at scale. He also mentioned the increasing demand for AI product leaders—individuals who possess a deep understanding of banking, IT, and analytics, and who can bridge the gap between these disciplines to ensure successful AI implementation.

Boteju added that AI should be seen as a tool to supplement human work, particularly in tasks that are too repetitive or complex for humans to handle efficiently. For example, AI can handle large-scale document analysis in KYC processes, freeing up human employees to focus on higher-value tasks, such as interpreting data and making strategic decisions.

It’s a really in depth and insightful conversation, with even more revealed within. Be sure to watch the full video above and check out more of our Virtual Arena panels on our website.