Head of Data at CLSA

Vincent Kwan

The financial sector, renowned for embracing technological advancements, is currently captivated by Generative AI's potential. Numerous experts have emerged, with firms, consultancies, and IT organizations positioning themselves as domain specialists. However, as past trends like blockchain have shown, hasty adoption without a thorough understanding can lead to unforeseen challenges.

Generative AI, celebrated for its ability to generate content from recognized patterns, is undeniably transformative. Yet, its current optimal use is as a supplementary tool rather than an infallible solution. While "Generative" implies innovation, it's essential to recognize that it doesn't ensure perfection. At its core, Generative AI, like most AI models, is based on statistical models. Despite its intricate design using billions of parameters, its predictions are more probabilistic in certain scenarios due to the nature of its training, introducing an inherent margin of error.

This margin of error in Generative AI differs from traditional software errors. While conventional programming issues, often traceable, arise from deterministic processes like incorrect specifications or flawed implementation, errors in Generative AI can emerge even when the system is functioning as designed due to its probabilistic predictions. Regulations such as the GDPR require that AI decision-making processes be transparent, accountable, and open to challenges to protect individuals' data rights. Researchers are still navigating the complexities of instilling explainability into AI outputs. For pivotal institutions like Investment Banks, introspection is vital: Is Generative AI, with its sometimes opaque decision-making process, genuinely ready for deployment?

To better understand the practical implications of Generative AI in investment banking, let's examine some specific applications:

• M&A Forecasting: Generative AI can create hypothetical merger scenarios based on past data, offering insights into potential M&A outcomes. However, while it provides quantitative insights, integrating qualitative factors such as organizational culture remains a challenge, potentially affecting its accuracy.

• Trade Optimization and Strategy Development: Using Generative AI to simulate past trades helps traders spot successful patterns and improve their strategies. While traders make the final decisions, insights from Generative AI offer them a data-driven perspective, aiding but not guaranteeing smarter investments.

• Enhanced Risk Assessment and Fraud Detection: Generative AI can simulate various market conditions, offering deep risk assessment insights. However, its challenge lies in events for which there is no historical data, potentially missing details that a banking expert would consider.

In investment banking, the potential of Generative AI is remarkable. It promises transformative changes, but addressing its intrinsic margin of error remains crucial. With diligent design and continuous improvements, this error and uncertainty can be significantly reduced e.g. by embarking better quality and quantity of data for training and fine-turning. When managed effectively, Generative AI can evolve from a supporting role to being a central player in banking processes. This progression doesn't overshadow the significance of human-AI collaboration; instead, it emphasizes that both AI and humans are essential for the future. Together, they herald a promising and advanced era in investment banking.