By CDO TMES contributing executive author and CIO, Weiyee In.
Introduction: The Allure and Pitfalls of Generative AI in Finance
The consumer level adoption surrounding generative AI, particularly free-use models like Generative Pre-trained Transformer (GPT) and its successors, has been significant in recent years, driven in large part by marketing hype and the broad applicability of Large Language Models (LLMs) and its ability to process and generate human language. While generative AI, especially LLM has demonstrated many remarkable capabilities, it is not in fact suitable for all manner of applications and solutions, especially within financial institutions. There are many reasons why banks and other financial institutions need to take a measured approach to adoption and implementation for AI and ML.
Regulatory Quagmire: Compliance in an Age of ‘Black Box’ AI
Financial institutions handle sensitive customer and business data, including financial transactions, personal information and intellectual property and need to ensure and demonstrate adherence to strict regulatory frameworks for cybersecurity, data security and privacy. The need to demonstrate that adherence creates a pressure on financial institutions for Transparency, Interpretability, Explainability and Auditability. Because so many of the AI offerings in the public domain are “black box” models demonstrating risk analysis and decision making to regulators becomes daunting. Even AI solutions that are purported to be “open” and “transparent” tend to have “non-trivial” complex architectures based on highly intricate neural network architectures with numerous layers and millions of parameters.
Legacy vs. Modern AI: The Feature Engineering Conundrum
Unlike “traditional or legacy” AI based on machine learning and supervised and iterative machine learning where an operator or a subject matter expert within the financial institution is able to not only understand the importance of features used for predictions, but can often configure and tweak or shape those features, with new AI models the model itself learns feature representations from the data, making it unclear which features are critical for decision-making. This puts the financial institution at significant risk for its ability to generalize from data and make predictions. Where historically in banks feature extraction has been performed by data scientists working with operational or business subject matter experts and their respective technology or engineering teams, with new AI raw data is ingested by the model, and the model “learns” to transform this data into meaningful representations.
Efficiency vs. Risk: The Double-Edged Sword of AI Automation
The benefits of new AI solutions is that they are efficient and can relieve human operators from the tasks of feature engineering, which tend to be extremely time-consuming, require domain expertise, technology and data science expertise, not to mention significant amounts of caffeine and eye drops. The painful work is alleviated because models can automatically capture relevant information from the data, but the risk is that the relevance and accuracy of that process depends on the model adapting to the variations, variability and vagaries of that data. In financial services, certain use cases, such as Long Short-Term Memory (LSTM) networks, a type of Recurrent Neural Network (RNN) architecture designed to capture and remember long-range dependencies in sequences are very useful for handwriting recognition and time-series data for credit scoring or fraud detection.
LSTM RNNs: A Tried-and-True Approach Facing Regulatory Scrutiny
The majority of the financial services industry globally currently employ LSTM RNN based solutions, however they have been struggling for over two decades in explaining LSTM RNNs to regulators. Over the past three decades financial institutions have leveraged deterministic AI (based on predefined rules, following very strict decision-making processes and is not adaptable or self-improving) and Reactive AI (also follows predefined rules, but it can handle multiple rules and adapt based on specific inputs or conditions) for a number of different applications. The broad majority of these use cases have historically been as support or augmentation to human operators for functions from automating more routine processes like document validation, account reconciliation, and compliance checks, where human intervention is the oversight layer for controls and anomaly detection. Reactive AI can make decisions based on programmed logic but doesn’t learn from data or draw conclusions and generalizations.
Emerging Trends: The Rise of Deterministic and Reactive AI
New AI solutions, especially Generative AI that can learn from data and generate new data, such as text, images, or predictions, based on patterns it has observed have suddenly become the rage because they appear to be able to adapt to different contexts and scenarios and thereby potentially could handle more complex and creative tasks. The main drawback or backlash within financial institutions have been computing and storage intensity and security. Setting aside the myriad of risk and security issues for any financial institution not following fundamental risk and security requirements associated with the use of generative AI and its potential impact on data sensitivity, security and privacy or the impact on the financial institution’s inherent risk profile there is a question of use case, application and maturity.
Generative AI: A Boon or a Burden for Financial Institutions?
One single technology rarely fits every solution requirement. Algorithmic trading for example, has long depended upon LSTM RNNs because the technology is specifically designed for handling time series data, making it a logical choice for algorithmic trading strategies that depend upon historical market data, technical analysis, price forecasting, predicting market volatility, etc. However LSTM is not as well suited for Natural Language Understanding (NLU) or even more basic Natural Language Processing for text analysis, language translation, analyzing non-time series data (such as news articles, “tweets” and other social media content, and financial reports) all useful for sentiment analysis for trading. Conversely generative AI such as LLMs are not suited for processing numeric data, or performing complex mathematical operations efficiently.
The CDO TIMES Bottom Line: A Holistic Strategy for AI Governance in Finance
It behooves financial institutions globally to understand the tools that can put their data to highest and best use and how, and in turn demonstrate and explain that to regulators to not only show transparency but also operational efficiency. The responsible and ethical use of AI, be it deterministic, reactive or generative requires a multi-faceted approach that combines an AI governance framework, security oversight, a data governance and data management plan that maps against organizational policies and compliance measures with technological controls and safeguards with continuous monitoring integrated to an incident response plan in place. As always a comprehensive risk assessment that takes into account the unique security challenges presented by any form of AI/ML and the means to implement appropriate controls to safeguard data, ensuring that AI systems adhere to the bank’s security policies and standards.
Love this article? Embrace the full potential and become an esteemed full access member, experiencing the exhilaration of unlimited access to captivating articles, exclusive non-public content, empowering hands-on guides, and transformative training material. Unleash your true potential today!
In this context, the expertise of CDO TIMES becomes indispensable for organizations striving to stay ahead in the digital transformation journey. Here are some compelling reasons to engage their experts:
- Deep Expertise: CDO TIMES has a team of experts with deep expertise in the field of Digital, Data and AI and its integration into business processes. This knowledge ensures that your organization can leverage digital and AI in the most optimal and innovative ways.
- Strategic Insight: Not only can the CDO TIMES team help develop a Digital & AI strategy, but they can also provide insights into how this strategy fits into your overall business model and objectives. They understand that every business is unique, and so should be its Digital & AI strategy.
- Future-Proofing: With CDO TIMES, organizations can ensure they are future-proofed against rapid technological changes. Their experts stay abreast of the latest AI advancements and can guide your organization to adapt and evolve as the technology does.
- Risk Management: Implementing a Digital & AI strategy is not without its risks. The CDO TIMES can help identify potential pitfalls and develop mitigation strategies, helping you avoid costly mistakes and ensuring a smooth transition.
- Competitive Advantage: Finally, by hiring CDO TIMES experts, you are investing in a competitive advantage. Their expertise can help you speed up your innovation processes, bring products to market faster, and stay ahead of your competitors.
By employing the expertise of CDO TIMES, organizations can navigate the complexities of digital innovation with greater confidence and foresight, setting themselves up for success in the rapidly evolving digital economy. The future is digital, and with CDO TIMES, you’ll be well-equipped to lead in this new frontier.
Do you need help with your digital transformation initiatives? We provide fractional CAIO, CDO, CISO and CIO services and have hand-selected partners and solutions to get you started!
Subscribe now for free and never miss out on digital insights delivered right to your inbox!