This semi-monthly column highlights news, government documents, NGO/IGO papers, industry white papers and reports, academic papers and speeches on the subject of AI’s fast paced impact on the banking and finance sectors. The chronological links provided are to the primary sources, and as available, indicate links to alternate free versions.
NEWS:
International Banker, September 11, 2024. The Imperative to Realizing AI’s Potential in Banking: Finding the Right Fit. Driven by its ability to automate tasks, personalize user experiences and manage risks, AI adoption is accelerating in banking and impacting how institutions operate and serve customers. AI can enhance decision-making, resulting in new insights and predictions; automate operations, improving speed and accuracy; and produce intelligent engagements with users, boosting customer experiences. Despite these possibilities, many corporate and retail banks may not be prepared to implement AI at scale across their businesses. This begs the question: Why would banks hesitate to adopt AI?
Forbes via MSN, September 5, 2024. Banks bet on a tech surge. It’s paying off big time as AI, software bring ‘sea change’ in productivity. Bankers are generating more revenue for their companies than ever. With the help of a steady stream of technological innovations—from online banking to artificial intelligence—the per capita contribution of bank workers has grown $98,000 since 2009. Put another way, financial firms would require 400,000 additional bankers to generate today’s levels of revenue had productivity remained unchanged from 15 years ago. While all of this has boosted the banks’ bottom lines, the news is less rosy for those who work in the banking industry. Even as the number of working Americans has increased 7.8%, the number of jobs in banking has stayed roughly constant at 1.37 million—meaning there’s 5.4% fewer banking jobs per American worker than there was 15 years ago.
Harvard Business School (HBS) has published a case study on DBS’ strategy towards Artificial Intelligence (AI), highlighting the bank’s use of AI. Developed over the course of eight months, the case is the first relating to AI that HBS has done on an Asian bank, and the first for a Singaporean company. Authored by Professor Feng Zhu, MBA Class of 1958 Professor of Business Administration at HBS and Co-Chair of the Harvard Business Analytics Program, the case maps the bank’s strategy and implementation as it industrialised its use of AI since 2014 to unlock business value, as well as how DBS is now approaching Generative AI. Professor Zhu has authored over 80 articles, cases, and notes in prestigious practitioner and academic journals, including the Harvard Business Review, American Economic Review, and Management Science. His research has also won international awards, such as the Inaugural Practical Impacts Award from the INFORMS Information Systems Society, which recognises business school academics with outstanding leadership and sustained industry impact through their research. The case study can be accessed at: https://www.hbs.edu/faculty/Pages/item.aspx?num=66332
GOVERNMENT DOCUMENTS:
Artificial Intelligence: Agencies Are Implementing Management and Personnel Requirements. GAO-24-107332 Published: Sep 09, 2024. Publicly Released: Sep 09, 2024. Artificial intelligence is rapidly changing the world and could improve government operations. For example, federal agencies are already using AI to analyze weather hazards. In October 2023, an Executive Order was issued to guide a coordinated approach to safely developing and using AI in government. It includes over 100 requirements with eight guiding principles like advancing equity and protecting privacy. We looked at 13 requirements of the Executive Order with clear expectations for what agencies should have implemented by March 2024. Agencies fully carried out these 13 requirements, laying the groundwork for government-wide AI efforts.
Digital Surveillance of Workers: Tools, Uses, and Stakeholder PerspectivesGAO-24-107639 Published: Aug 28, 2024. Publicly Released: Sep 11, 2024. In 2023, the White House Office of Science and Technology Policy asked for public comments on employers’ use of digital surveillance to monitor workers’ activities. We reviewed the 217 comments from workers, unions, tech developers, and others. Commentors noted that cameras and monitoring software are frequently used to track productivity, performance, safety and health, and security. They offered differing views on how these tools affect workers. Some trade associations said that they increase productivity and prevent injuries and illnesses. However, some workers and unions said that they increase stress and negatively affect morale.
Bolesta, Karolina and Akar, Mutlu and Coita, Ioana and tarantola, claudia and Iannario, Maria and Osterrieder, Joerg and Sipos, Ciprian and Schwendner, Peter and Bedowska-Sojka, Barbara and Pisoni, Galena and Maxhelaku, Armela and Maxhelaku, Suela and Weinberg, Abraham Itzhak and arakelian, veni and Rupeika-Apoga, Ramona and Giordano, Sabrina and Filipovska, Olivija and Gomez Teijeiro, Lucia and Bernard, Frédérik Sinan, AI-Driven Failed Trials in Investment Strategies: A Network Data Analysis Approach (September 02, 2024). Available at SSRN: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4944243. In recent years, the intersection of ArtificialIntelligence (AI) and quantitative finance has sparked significant interest for formulating and guiding investment strategies. In contrast to the leading discourse focusing on AI success case studies, this paper addresses particularly “failed trials” driven by AI implementations for investment strategies and on the strategic use of AI to simulate and learn from such failures. Understanding the underlying factors that lead to under-performing AI-powered solutions for investment and the parameters used in AI simulations of failed trials is instrumental to guide future developments towards designing more resilient AI systems for investment. In this context, we introduce network data analysis as a powerful tool to enhance these models by capturing complex interdependencies and systemic risks within financial markets. Our study also addresses the broader implications of explainable AI and policy frameworks for AI-powered investment, emphasizing the need for transparency in finance AI-driven decision-making. Together, this paper proposes integrating advanced AI methodologies with network data analysis, while emphasizing explainability and policy orientation, therefore contributing holistically to both the academic discourse and practical applications of these technologies in risk management and investment optimization.
NBER – The Impact of Cloud Computing and AI on Industry Dynamics and ConcentrationYao Lu, Gordon M. Phillips & Jia Yang Working Paper 32811. DOI 10.3386/w32811. Issue Date We examine the rise of cloud computing and AI in China and their impacts on industry dynamics after the shock to the cost of Internet-based computing power and services. We find that cloud computing is associated with an increase in firm entry, exit and the likelihood of M&A in industries that depend more on cloud infrastructure. Conversely, AI adoption has no impact on entry but reduces the likelihood of exit and M&A. Firm size plays a crucial role in these dynamics: cloud computing increases exit rates across all firms, while larger firms benefit from AI, experiencing reduced exit rates. Cloud computing decreases industry concentration but AI increases concentration. On the financing side, firms exposed to cloud computing increase equity and venture capital financing, while only large firms increase equity financing when exposed to AI.
Hussain, Safdar and Bharathy, Gnana and Aziz, Saqib, Explainable Artificial Intelligence in Financial Services: A Case Study on Credit Card Delinquency (September 01, 2023). Available at SSRN: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4930148. This study delves into explainablity of an internally built credit-risk algorithm that has been specifically designed to forecast credit card delinquency. In accordance with the principles of Explainable ArtificialIntelligence (XAI), this research sets out to explain the inner workings of learning models and provide light on the reasoning behind certain judgments. This case study uses advanced machine learning algorithms, notably XGBoost, to construct a predictive framework by utilizing financial data for the year 2022. The integration of sophisticated XAI techniques, such as SHAP, LIME, PDP, ICE, and counterfactual explanations, plays a pivotal role in enhancing the levels of transparency and explainablity. The merging of predictive precision and actionable insights through XAI facilitates the establishment of a pioneering paradigm for credit risk assessment models. This integration facilitates the process of making well-informed decisions in the financial domain. The importance of combining predictive accuracy with transparent explanations is underscored through rigorous analysis, leading to robust and conscientious credit risk management techniques that exert influence across financial service industries. Further this study, showcases the performance measure “Explanation Fit” that measures the “Goodness of Explanation” by one of the XAI technique-LIME framework. Within this narrative, the exploration of these paradigms affords a panoramic vista, enriching our understanding of their pragmatic applications and reinforcing the mandate of transparency and explainablity.
Raza, Asif, Empowering Global Banking Through AI-Driven Risk Management: A Practical Framework for Optimization and Methodological Integration (August 27, 2024). Available at SSRN: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4938239 In the current global banking world, AI-driven risk management is the future of risk management and the solution to many challenges affecting the banking sector. This essay explores the practical framework for optimization and methodological integration of AI in risk management in global banking. Financial institutions have found a lasting solution to major risk management problems by integrating AI-driven techniques. The paper also discusses the theoretical foundations of AI optimization techniques and methodological integration. The paper provides real-world examples and actionable recommendations that can guide banking and financial institutions in leveraging AI to strengthen risk management.
Bennett-Lovsey, Robert, In an era of big data and artificial intelligence, how do group dynamics and emotional responses influence the adoption of InvestTech among institutional investors, and how do these factors continue to shape its use in the investment decision-making process? (March 05, 2024). Available at SSRN: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4937888. In the complex landscape of institutional investing, the emergence of big data and artificialintelligence (‘AI’) has ushered in an era of transformative change. This change is epitomised by the rapid evolution of sophisticated financial technology systems that harness AI and extensive data analytics to aid investment decision-making both directly and indirectly (‘InvestTech’). Through primary and secondary sourced interview data, this thesis aimed to explore the influence of group dynamics and emotional responses on the adoption and ongoing use of InvestTech among institutional investors (‘investors’). Thematic analysis was applied to the interview data and revealed three fundamental challenges faced by investors, namely, (i) understanding the context, (ii) dealing with uncertainty, and (iii) communicating under ambiguity. It also identified strategies investors use to navigate these challenges and explored collective emotional responses to InvestTech, such as emotional ambivalence. To gain further understanding, a systems psychodynamics approach was applied in conjunction with concepts from sociotechnical systems theory and complexity theory. The research identified social defences as a consolidated theme and underscored the need for investors to address social and human factors, as well as unconscious drivers during complex adaptive challenges, such as InvestTech implementation and use. Furthermore, the study examined the role of emotional contagion in amplifying individual feelings to the group level and influencing their behaviour, both during and after InvestTech adoption. Lastly, the findings suggest InvestTech can fundamentally impact human-to-human group dynamics, most notably in relation to accountability, responsibility, and potentially risk tolerance levels in the investment decision-making process.
