APPLIED GENERATIVE AI GOVERNANCE: A VIABLE MODEL THROUGH CONTROL AUTOMATION
GERHARDT SCRIVEN | MARCEL BRAGA | DIOGO SANTOS | DIEGO SARAI
GERHARDT SCRIVEN | Managing Principal, Capco
MARCEL BRAGA | Principal Consultant, Capco
DIOGO SANTOS | Principal Consultant, Capco
DIEGO SARAI | Managing Principal, Capco
Generative AI has the potential to revolutionize the banking industry with hyper-personalization and advanced chatbots. However, the technology also poses risks to trust, accuracy, fairness, intellectual property, and confidentiality that all need to be mitigated to ensure that the benefits of Generative AI are realized.
In this article, we explore practical considerations to help mitigate these risks through the construction of a governance framework that has a focus on AI explainability, intellectual property protection, and minimizing model hallucination. We then derive a control framework against these key outcomes and present technology solutions we built around automating some of the key controls towards making our governance model viable. Finally, we explore what other institutions are doing in the field of generative AI governance and discuss new emerging roles needed to execute against the governance model.
In terms of practical application, we recommend that financial institutions start small when it comes to generative AI governance and focus on defining a “minimum governance model” on a use case by use case basis to minimize the time and cost footprint of governance. We also recommend that governance is implemented very early in the solution lifecycle so that it is baked in at root-level; hence, reducing churn and rework of the solution when industrializing the use case within the financial institution.
IAN HUNT | Buy-Side Industry Consultant and Adviser
Digitalization is not about doing what we do now, but with slightly better technology. It is an opportunity to do something very different, which is much simpler, much cheaper, and much better. A number of factors are preventing us from taking, or even perceiving, that opportunity. We are held back by the popular perception that digital assets are questionable and shady, and by regulatory uncertainty over the treatment of digital assets, as well as by our own unwillingness to think beyond the current financial ecosystem: we find it hard to accommodate the degree of change that would enable us to
maximize the benefits of digitalization.
This paper explains the radical potential of native digital assets to create a single, simple issuance and transaction model across all financial assets. This would deliver a dramatic improvement in client outcomes and in the flexibility of investment products, along with an equally dramatic reduction in cost and risk in the way that we deliver those products. It would enable regulation to be far simpler, but more effective. It would allow us to roll back the surging tide of complexity in the infrastructure, management, and regulation of finance.
This model will be adopted soon in a forward-thinking jurisdiction. All other markets, ploughing on with our current overweight, over-complex, and heavily regulated financial ecosystem, will become uncompetitive: they will have no choice but to follow.
JESÚS LOZANO BELIO | Senior Manager, Digital Regulation, Regulation and Internal Control, BBVA
The rapid advancement of artificial intelligence has facilitated the automation of previously challenging tasks. This article
explores the opportunities and benefits associated with AI adoption, specifically within the banking sector. It examines how banks are currently utilizing AI, the challenges they face in implementing AI systems, and the role of regulators in supporting AI adoption. Additionally, as the article has been written with the help of some AI tools, it serves as a practical demonstration of AI’s applicability in research and information dissemination. While AI demonstrates proficiency in these areas, it is important to note that human expertise and supervision remain essential due to inherent limitations of the technology.
ANNA OMARINI | Tenured Researcher, Department of Finance, Bocconi University
Technology in banking has always had the power to affect the fundamentals of business, such as information and risk analysis, distribution, monitoring, and processing. The relationship between technology and banking is, however, quite different to how it used to be, predominantly due to stronger interdependencies, both technological as well as strategic. Today’s digital technologies have the power to improve efficiency and effectiveness in services, as well as exerting increasing influence on banks’ products and delivery methods, and increasingly on strategies.
Digitalization is changing the rules of the game in many industries, and this results in the emergence of complex and dynamic ecosystems for growth and innovation. The main forces shaping these changes have led the financial services industry to reconsider the role of banking and finance, more as an “enabler” for many other businesses and commercial initiatives than as a mere provider of products and services.
This paper looks at how financial services organizations are transforming themselves using the new technologies at their disposal and tries to determine what should be kept and what needs to change.
JESSICA TAYLOR | Consultant, Capco
IVO VLAEV | Professor of Behavioral Science, Warwick Business School
ANTONY ELLIOTT OBE | Founder, The Fairbanking Foundation
The FCA’s Consumer Duty regulation aims to transform financial services for customers by requiring firms to consider the needs, characteristics, and objectives of all their customers, and how they behave, at every stage of the customer journey.
Its success, however, is dependent on compliance from firms and with new regulations, there often exists a policy implementation gap whereby policies do not lead to changes in behavior.
This study provides a novel approach by applying “behavioral science frameworks” to compliance with financial regulation, improving outcomes for customers under the FCA’s Consumer Duty and future financial regulatory change.
PRZEMEK DE SKUBA | Senior Consultant, Capco
BIANCA GABELLINI | Consultant, Capco
JESSICA TAYLOR | Consultant, Capco
Customer vulnerability is one of the key concerns of the Consumer Duty regulation, a very welcome ESG-aligned enhancement of financial institutions’ governance. Adherence to the regulation requires a clear focus on data collection that helps lenders manage the impact of consumer vulnerabilities without imposing penalties or resulting in a negative
impact on clients.
There are two parts of the problem that need to be addressed: firstly, how to capture vulnerability data by encouraging clients/consumers to voluntarily submit the information (the behavioral aspect) and secondly, how to technically capture, manage, and store this data to ensure compliance with the Consumer Duty regulation.
This article considers both problems and reviews the tools from behavioral science that can encourage customer disclosure and two key technology solutions (data lakes and blockchain) to comply with the capture, management, and storage of data whilst remaining GDPR compliant and fully aligned to the objective of voluntary submission of information regarding vulnerabilities by clients/consumers.
FAYSSAL MERIMI | Managing Principal, Capco
JULIEN KOKOCINSKI | Partner, Capco
The concept of “developers 3.0” is emerging, defining the new avant-garde generation of software development professionals. These developers, transcending traditional skills, place generative artificial intelligence at the heart of their approach, thus revolutionizing software design and development paradigms.
This article explores the methodologies and strategies adopted by these innovators, highlighting notable advantages in terms of productivity and quality. At the same time, we address the challenges associated with this combination of traditional software development practices with the new methodologies centered around generative artificial intelligence, such as ethical issues, security concerns, and the need to maintain a balance with traditional skills.
Our analysis aims to provide an in-depth perspective on the growing influence of generative AI in the field of software development and its implications for the future of the profession.
DENNIS VETTERLING | Doctoral candidate, Institute of Information Management, University of St. Gallen
ULRIKE BAUMÖL | Executive Director of Executive Master of Business Administration in Business Engineering, and Senior Lecturer on Business Transformation, University of St. Gallen
Business today is not conducted by single organizations alone but in networked designs with diverse actors. A construct where actors engage in joint value creation is called a business ecosystem. Specifically, within the context of core services originating from the financial services industry, such constructs are called financial business ecosystems. Innovative technologies and intelligent methods enable value creation in these organizational set-ups. To effectively participate in these ecosystems and exploit the potential of innovative technologies and intelligent methods, organizations need to
develop a novel operating model.
We propose a blueprint for such an operating model building on two levels of capabilities: first level capabilities that enable the exploitation of data and the number of partner relations as underlying resources of business ecosystems. The proposed second level capabilities enable the organization to engage in business ecosystems. By suggesting these capabilities, we aim to guide organizations on a targeted transformation journey and enable them to leverage innovative technology for actively engaging in financial business ecosystems.
ALI HIRSA | Professor of Professional Practice, Department of Industrial Engineering and Operations Research, Columbia University, and Chief Scientific Officer, ASK2.AI
SATYAN MALHOTRA | Chief Executive Officer, ASK2.AI
Artificial intelligence is a very powerful application whose time has come. At a quick glance, it can be really seductive to
believe, for example, the purveyors of xxxGPT, that its deployment is as simple as pushing a button or is a “data in, miracles out” strategy. However, harnessing it effectively requires navigating a myriad of options embedded within its critical pillars of data, models, and visuals. The complexity is accentuated by the deployer’s capabilities and the organization’s openness to change, as outcomes move from rules to an objective-based spectrum. In navigating these challenges lies the key to
W. PAUL CHIOU | Associate Teaching Professor of Finance, Northeastern University
YUCHEN DONG | Senior Engineer, MathWorks
SOFIA X. MA | Senior Engineer, MathWorks
Applying machine learning techniques to improve the accuracy and efficiency of predictions of credit risk rating is increasingly critical to the financial services industry. In this study, we apply MATLAB to investigate the performance of two approaches, decision forest and boosting algorithms, by using a wide range of financial data. The empirical outcomes suggest that both methods exhibit considerable performance but may be superior to each other in different scenarios. Boosting algorithms method exhibits an accuracy rate of approximately 67% across the credit rating categories. The random forests model generates lower accuracy rates for low and medium classifications than the boosting method, but the accuracy rate for high credit ratings reaches 79%, more accurate than the boosting method.
NIRAN SUBRAMANIAM | Associate Professor in Financial Management & Systems, Henley Business School
Digital transformation revolutionizes how businesses provide value by seamlessly integrating digital technologies into operations, strategies, and culture. Its core objectives encompass enhanced efficiency, elevated customer experiences, and heightened competitiveness, while ensuring adaptability in the face of swiftly evolving technology and market landscapes.
A key enabler in this transformation is artificial intelligence, which infuses intelligence and automation into digital technology utilization. AI’s capabilities encompass mining and analyzing diverse organizational data to unearth patterns that drive recommendations and inferences. For instance, customer data analysis unveils preferences, enabling personalized marketing and lucrative opportunities such as cross-selling and up-selling. AI, with its pattern recognition, inference, recommendation, and predictive analytics, is at the forefront of driving digital transformation in organizations.
This article proposes a framework for successful digital transformation in organizations.
ALNOOR BHIMANI | Professor of Management Accounting and Director of the South Asia Centre, London School of Economics
Digital transformations are taking place across enterprises in every industry. Becoming digital is both essential to compete and virtually unstoppable. All previous major technological disruptions have led to financial intelligence being altered to ensure more effective decision making in the face of change.
This article considers issues that organizations going digital need to address in relation to accounting information provision. It discusses several points: accounting’s need to move toward the delivery of predictive information rather than relying on extrapolations of historical data; the recognition that machines make more decisions that alter accounting information needs, structures, and contents; the importance of recognizing the “data-learning-action” loop that is emerging; the emergence of “strat-perational” information contexts; and the relevance of prioritizing qualitative insights in decision making.
CLAUDE DIDERICH | Managing Director, innovate.d
Artificial intelligence can be considered one of those technologies, like 5G, 3D printing, and virtual reality, that can disrupt the business world. While AI has the potential to solve meaningful business problems, implementing it in a way that creates value is challenging. Unfortunately, many AI proponents lack the necessary computer science and mathematics machine learning skills required for developing AI systems that pass the Turing test.
This paper presents an assessment of the characteristics of AI, allowing the reader to understand what specific business problems it can solve, and describes how an AI-supported investment advice solution for wealthy private clients can successfully deliver value. By reviewing the lessons learned, I conclude that the future of AI is bright if the focus is put on applying it to those challenges that it is best suited to solve.
UDO MILKAU | Digital Counsellor
Since the launch of the generative artificial intelligence tool ChatGPT end of 2022, there has been an incredible public awareness. Doomers predicted an end to humanity, while more reasonable assessments discussed the impact on traditional industries and on the workforce.
In a nutshell, generative artificial intelligence is nothing more than statistical estimation and continuation of an input sequence based on a text corpus of the past. To evaluate the actual impact of generative artificial intelligence (AI) and large language models (LLM), this paper uses the case of asset management as a benchmark. These statistical estimators can produce a “next best token” based on an “internet average”, i.e., tremendous text corpora gathered from internet sources, but which cannot understand, predict anything new, or create something innovative. Consequently, generative AI/LMMs can augment staff to perform “on average”, or help internet users obtain an “average answer” to their questions about financial management.
While this can (and probably will) change the future landscape of financial advice and the way consumers access information, generative AIs/LMMs are far from any type of “superintelligence”. The potential danger of misuse by human actors, however, remains the biggest danger and has to be monitored closely.
MARIA MOLONEY | Senior Researcher and Consultant, PrivacyEngine, Adjunct Research Fellow, School of Computer Science, University College Dublin
EKATERINA SVETLOVA | Associate Professor, University of Twente
CAL MUCKLEY | Professor of Operational Risk in the Banking and Finance Area, UCD College of Business, and Fellow, UCD Geary Institute
ELEFTHERIA G. PASCHALIDOU | Ph.D. Candidate, School of Economics, Aristotle University of Thessaloniki
IOANA COITA | Consultant Researcher, Faculty of Economics, University of Oradea
VALERIO POTI | Professor of Finance, Business School, University College Dublin, and Director, UCD Smurfit Centre for Doctoral Research
The proliferation of artificial intelligence is reshaping modern life in many ways. This has prompted action from many governments globally. The European Union is in the process of drafting a new E.U. AI Act, modelled on GDPR. To navigate this evolving regulatory landscape, financial researchers and industry professionals will need comprehensive training. However, existing efforts seem limited.
This paper puts forth the idea of tailored training to better understand the complex interaction of data protection and ethical AI. It uses case studies to highlight the challenges of AI and the GDPR in the financial services sector. We also put forth survey findings that suggest current programs inadequately prepare individuals for GDPR compliance in AI. Recommendations include an initial training framework for ethical and compliant AI engagement.