Dismantling the Zeal and the Hype: The Real GenAI Use Cases in Risk Management

Dismantling the zeal and the hype – the real GenAI use cases in risk management As firms pour money into GenAI and commentators raise the age-old specter of hype, clarity on what’s actually useful is lacking. LLMs should be viewed within the broader AI ecosystem, while references to hype cycles tend to disregard the wider societal forces that shape the development of technology. In this clear-headed view of the field, Chartis explores the advantages and drawbacks of GenAI applications in risk management – firmly within the well-established and continuously evolving AI landscape. Overview: more than words In this Market Insight report, Chartis considers the current boom in artificial intelligence (AI) and the periods of heightened interest and investment in the field that preceded it. We also outline the potential applications in financial services that are emerging from the new wave of generative (Gen) AI, and suggest the use cases that are most likely to succeed. Chartis argues that although a proportion of hype-driven GenAI use cases in financial risk management may not mature, the deep learning methods and emerging infrastructure that underpin them represent important and enduring innovations. The model architectures that are powering the current hype in GenAI and large language models (LLMs) are defining a new era of sophisticated natural language processing (NLP) and document transparency in financial services. We suggest that a golden era of NLP is emerging, in which text or ‘text-like’ information (such as programming languages) can be flexibly processed, parsed and generated. (NLP applications include text search, topic modeling, sentiment analysis, extraction, summarization and generation.) In addition to text generation and code development and optimization, another GenAI use case the ability to generate synthetic data, although this area is receiving comparatively less attention. Within the broader GenAI space, this report will focus largely on the application of LLMs in risk management. We also consider the impact that developments in GenAI are having on the uptake of and attitudes toward array and vector-based databases in risk management. We also consider the broader implications of the popularity of vector databases on the already complex enterprise data stack, as well as the costs associated with implementing AI infrastructure for LLM use cases. Another key area of rapid growth we will consider is the broader AI ecosystem, including open-source tools and projects and hardware infrastructure. The computationally intensive process of training and deploying LLMs at scale requires domain-specific hardware, infrastructure software and vector databases, and is thereby driving a boom in AI infrastructure.