According to research by The Alan Turing Institute, large language models (LLMs) have the potential to increase efficiency and safety in the finance sector by identifying fraud, producing financial insights, and automating customer care.
There is growing awareness of the potential to improve services across a range of sectors including healthcare, law, education, and financial services including banking, insurance, and financial planning because LLMs can analyze large amounts of data quickly and produce coherent text.
This report shows that people working in this field have already started using LLMs to support a variety of internal processes, like the review of regulations, and are evaluating its potential to support external activity like the delivery of advisory and trading services. It is the first to explore the adoption of LLMs across the finance ecosystem.
43 professionals from major high street and investment banks, regulators, insurers, payment service providers, the government, and the legal profession participated in a workshop conducted by researchers in addition to a survey of the literature.
52% of workshop attendees are already using these models to improve their performance in information-oriented tasks, such as managing meeting notes, cyber security, and compliance insight. 29% of participants use them to strengthen their critical thinking abilities, and 16% use them to simplify difficult tasks.
Additionally, the industry is already setting up technologies to increase efficiency by quickly analyzing vast amounts of text in order to streamline risk profiling, investment research, and back-office activities.
Participants in a survey regarding the future of LLMs in the finance industry predicted that, in the next two years, LLMs would be included into services like venture capital strategy development and investment banking.
Additionally, they believed that LLMs will probably be connected to enhance human-machine interactions. For instance, dictation and embedded AI assistants might simplify knowledge-intensive jobs like reviewing rules.
However, participants also agreed that there are risks associated with the technology, which would restrict its use. Financial institutions are restricted in their use of AI systems that they are unable to explain or that do not produce results reliably, consistently, or error-free due to a plethora of regulatory requirements and obligations.
The authors advise financial services professionals, regulators, and policy makers to work together across the industry to exchange and expand information regarding adopting and using LLMs, especially with regard to safety concerns, in light of their results. Additionally, they recommend that the increasing popularity of open-source models be investigated and that it might be properly used and maintained, with the top priority being the mitigation of security and privacy problems.
“Banks and other financial institutions have always been quick to adopt new technologies to make their operations more efficient, and the emergence of LLMs is no different,” stated Professor Carsten Maple, main author and Turing Fellow at The Alan Turing Institute. Through the collaboration of specialists from several fields within the banking industry, we have established a shared comprehension of the applications, hazards, advantages, and schedule for executing these technologies on a large scale.
“It’s really positive that the financial sector is benefiting from the emergence of large language models, and their implementation into this highly regulated sector has the potential to provide best practices for other sectors,” stated Professor Lukasz Szpruch, program director for Finance and Economics at The Alan Turing Institute. This study highlights the value of collaboration between academic institutions and business to evaluate the enormous potential and the pragmatic and moral difficulties of new technologies in order to guarantee their safe use.