New Ideas, Real Insights, Bold Perspectives

Kristof Meganck, Chief Data Officer, BNP Paribas Fortis
Bart Claeys, Belgium Data Practice Lead, Projective Group

The rapid acceleration of Artificial Intelligence (AI) adoption in financial services has shifted the focus from experimentation to industrialization. Banks are making tangible progress, with AI increasingly embedded in client-facing, operational, and compliance-related processes. At the same time, moving from successful use-cases to sustained, enterprise-wide adoption remains a complex, ongoing journey. As AI capabilities mature, underlying data foundations rather than algorithms alone emerge as the decisive factor shaping scalability, trust, and regulatory readiness. Drawing on industry research and practitioner experience from a large European bank, this paper examines how challenges related to data quality, ownership, lineage, accessibility, and governance intensify as AI moves into production. It highlights why traditional centralized and hybrid data models struggle to support this next phase, and how data mesh principles offer a pragmatic operating model shift. By treating data as a product, aligning ownership with business domains, and embedding federated governance, data mesh helps create the structural conditions required for scalable and compliant AI. The central conclusion is that progress in AI adoption is inseparable from progress in data maturity, making data management a strategic concern for financial institutions navigating the next phase of AI-driven transformation.

Budha Bhattacharya, Head of Systematic Research, Lombard Odier Investment Management, and Industrial Professor of Banking and Finance, Institute of Finance and Technology, University College London (UCL)
Francesca Medda, Professor of Applied Economics and Finance, and Director, Institute of Finance and Technology, University College London (UCL)

Environmental, social, and governance (ESG) integration in structured credit is becoming a practical necessity, yet most established ESG scoring methods were designed for single-issuer instruments and often break down when applied to securitizations. This article proposes a hybrid, natural language processing (NLP)-driven ESG scoring methodology tailored to structured credit particularly commercial mortgage backed securities (CMBS) where risk is distributed across multiple entities (originators, underwriters, servicers, trustees, and borrowers) and where critical evidence is embedded in large, unstructured transaction documents. The approach combines (i) advanced NLP for entity and role extraction from CMBS documentation, (ii) structured ESG metrics from reputable quantitative datasets, and (iii) news-based sentiment and controversy signals to improve responsiveness and reduce overreliance on self-disclosed information. The methodology also introduces hierarchical (role-based) weighting to reflect the fact that different parties exert varying influence over ESG outcomes and risk transmission. A Freddie Mac Green Bond CMBS case study illustrates how the approach can be applied in practice using document extraction, identifier linkage, and illustrative (hypothetical) quantitative and sentiment scores to demonstrate score decomposition and interpretability.

Richard P. Padbury, Director of Research and Ecosystems, BMO Financial Group
Ryan B. Duffy, U.S. Chief Data and Analytics Officer, BMO Financial Group
Kristin L. Milchanowski, Chief Artificial Intelligence (AI) and Data Officer, BMO Financial Group

In financial services, the transition from Machine Learning to goal-oriented autonomous AI Agents promises transformative improvements in speed, scale, and cost. Yet reliability at production scale is constrained by challenges rooted in the data foundation. Notably, hallucinations and decision errors are not generated from algorithmic model randomness, but from data drift, weak lineage, and stale or ungoverned information and context. This paper asserts that dependable AI is fundamentally a data-layer challenge and examines these challenges in the context of the banking sector. Substantive progress in Generative Artificial Intelligence (GenAI) results from strengthening a unified, governed, and observable data plane that embeds continuous data-quality telemetry, automated source-to-agent lineage, near real-time schema-change detection, and policy-as-metadata to enforce privacy and purpose-bound access. The architecture follows FAIR principles and supervisory expectations (e.g., SR 11-7/OCC 2011-12, BCBS 239, and GDPR/PIPEDA), and describes multi-agent integrity patterns (i.e., builder/checker) and self-healing pipelines that prevent silent error propagation. When combined, these controls and capabilities elevate the horizontal foundation to regulatory-grade standards and enable workflow-embedded agents in vertical functions (e.g., onboarding, lending, AML, and Treasury) to plan, act, and adapt with confidence while remaining human-centered in their design and application.

Gary Paul, Partner and U.K. Data Practice Lead, Projective Group
Keeley Miles, Chief Financial Officer, FCE Bank PLC

This article argues that data culture constitutes a foundational organizational capability one that AI cannot replace. It explores the concept of data culture in depth, examines its strategic importance within financial services, analyses the barriers to its development, and proposes practical mechanisms for cultivating and sustaining it. Drawing on academic literature, regulatory guidance, and industry practice, the article demonstrates that data culture is central to effective decision-making, innovation, trust, and long-term resilience.

Shahina Khan, EMEA Chief Data and Analytics Officer, SMBC Group

This article examines the evolving role of data governance in financial services amid rapid AI advancement and intensifying regulation. It highlights the dual dynamic of AI for data governance, where AI automates discovery, quality monitoring, lineage, and metadata enrichment, and data governance for AI, where trusted, well-governed data underpins compliant and explainable AI systems. Driven by frameworks such as BCBS 239, GDPR, DORA, NIS2, and the E.U. AI Act, institutions must strengthen metadata, ownership, and lineage across complex, multi-vendor environments. The paper proposes practical, phased strategies, promotes governed data products, and envisions AI-enabled governance ecosystems that deliver regulatory compliance, operational resilience, and scalable, trustworthy AI innovation.

Martijn Dekker, Professor of Business and Cybersecurity, University of Amsterdam, and CISO, ABN AMRO Bank N.V.

We live in an increasingly digital, complex, and volatile world. Organizations need to be able to prevent, absorb and respond to shocks caused by disruptions of digital services. These are, for example, due to ransomware attacks on suppliers or unavailability of cloud services. In this article, we will explain that the required digital resiliency is not only a technical problem, but a leadership and decision-making topic and hence requires new or improved governances. Cyber resilience is not only about controls, but about choices and how we make them. We conclude with providing practical guidance to improve the leadership of cyber resilience.

Laure Molinier, Director, Group Operational Resilience, Euroclear

Modern organizations face a rapid shift from traditional, infrastructure focused resilience toward data centric resilience, driven by evolving cyber threats, increasingly complex hybrid technology stacks, and heightened regulatory expectations. Cyberattacks, including ransomware, large scale data corruption, and hybrid warfare now target data directly, posing greater systemic risk than historical physical disruptions. This article explores this evolution based on the experience of Euroclear, a globally systemic Financial Market Infrastructure that safeguards €40.7 trillion in assets. Given the magnitude of assets protected, any disruption could have global repercussions, making perfect data and robust recovery strategies essential.

Tensie Whelan, Distinguished Professor of Practice Emerita, Business & Society, Stern School of Business, New York University

This paper highlights the inherent tension in environmental, social, and governance (ESG) reporting: ESG reporting metrics generally do not measure outcomes, they measure outputs. Outputs, such as a policy on a sustainability topic, do not necessarily drive performance, societal or financial. This article reviews the Sustainability Accounting Standards Board (SASB) and European Sustainability Reporting Standards (ESRS) standards and metrics through the lens of driving potential societal and financial performance and makes recommendations for the standards and corporate leaders.

Johan Gerber, Executive Vice President and Head of Security Solutions, Mastercard
Kate Pohl, Global Sales Lead, Projective Group

Challenging tomorrow’s cybercriminals requires intelligent, adaptive, and collaborative tools. Revolutionary technologies such as artificial intelligence (AI), and its many derivatives generative AI (GenAI) and Agentic AI, and quantum computing will provide the foot soldiers for battles no longer fought with static rules and blacklists, but through dynamic, real-time analysis and intervention. Sadly, these same technologies are also aiding cybercriminals in their efforts to undertake fraud on a larger scale, and more globally. This is an arms race in which threat and response evolve in tandem and in tension. In this article, we explain how challenging tomorrow’s cybercriminals requires not only the application of the latest and most advanced technologies but also collaboration among different parties involved in the Payments value chain, including among competitors. We further describe how Mastercard has been able to develop solutions that meet the challenges of managing, and mitigating, Payments fraud in today’s highly complex technological environment while remaining compliant with the myriad of global regulations.

Judith Arnal, Senior Research Fellow, Centre for European Policy Studies (CEPS), Elcano Royal Institute, and the European Credit Research Institute (ECRI)

Big Tech firms have expanded rapidly into financial services, integrating payments, wallets, and data-driven offerings into broader digital ecosystems. This article examines the economic logic behind that expansion and the resulting policy challenges for ensuring a level playing field. It first explains why Big Tech emerged in highly concentrated digital markets, driven by network effects, economies of scale and scope, and data feedback loops. It then analyses why financial services are a natural extension of platform models and maps Big Tech participation across the financial value chain in the U.S., Europe, and China. The paper highlights a key trade-off: Big Tech can enhance competition, innovation, and inclusion in the short term, yet may reinforce concentration and distort incentives over time. It proposes policy design principles and applies them to the E.U.’s FiDA debate on DMA-designated gatekeepers.

James Halcomb, Chief Research and Development Officer, EDM Association
Scott Beange, Head of Data Management and Cyber Strategy, Projective Group

With the number of organizations attempting to integrate Artificial Intelligence (AI) within their operations increasing, data risk which encompasses data quality, availability, governance, and security has been transformed from a localized IT concern into a strategic enterprise challenge requiring a board-level offensive. In this article, we will analyze the business implications of data risk, including financial losses, regulatory consequences, operational disruptions, reputational damage, loss of resilience and trust, and personal accountability for executives. We will also discuss the key steps that should be considered, from a strategic holistic perspective, in order to address the rapidly changing data risk challenges faced by organizations.

Vijay Varadharajan, Emeritus Professor, The University of Newcastle, and former Microsoft 
Chair Professor of Innovation, Macquarie University Sydney, Australia

In the current technology landscape, data is central to digital transformation as it drives decisions, enables personalization, and opens innovations. In this article, we explore how data centric approach is critical for secure data driven decision making. We present a data centric approach, where data will be a fundamental tenet in the specification of systems, services, and policies. We illustrate how such an approach provides superior security capabilities, by enabling the data owners to have better control over their personal data as the data moves over the Internet and is being used on distributed web platforms and network services. Systems that adopt a data centric approach are also more amenable to the enforcement of privacy regulations.

Bill Oates, Cloud and AI Capability Lead, Projective Group

As of 2026, the global financial services industry (FSI) remains entangled in a resurfaced Solow Paradox, where unprecedented investment in generative AI (GenAI) infrastructure has yet to yield significant increases in Total Factor Productivity (TFP). This article investigates the measurement gap by synthesizing Carlota Perez’s theory of technological revolutions with Sangeet Paul Choudary’s “Reshuffle” framework. Drawing historical parallels to the transition from group drive to unit drive systems during the electrification of industry supports the idea that AI’s economic impact is currently experiencing an adoption J-curve and suffering socio-institutional lag. By examining three application areas for AI in financial services, it is possible to see this lag directly and position the situation in 2026 as a turning point.

Xuan Zhang, Ph.D. Student in Computer Science, New Jersey Institute of Technology
Guiling Wang, Distinguished Professor of Computer Science, New Jersey Institute of Technology

This paper examines how AI and psychological principles converge in customer retention, integrating data-driven systems with behavioral insights to enhance loyalty and lifetime value. Through theoretical analysis and case studies, we identify integration points where AI leverages psychological principles and misalignment risks where optimization undermines authentic connection. Three critical conclusions emerge. First, sustainable retention requires AI systems governed by psychological guardrails, preventing optimization from eroding trust. Second, organizations must leverage AI for scalable pattern recognition while maintaining human oversight in emotionally sensitive interactions. Third, competitive advantage arises from orchestrating artificial and human intelligence, where algorithms reinforce rather than replace psychological connection. The future of customer retention depends on intentionally designed AI–psychology systems grounded in transparency, ethical constraint, and strategic collaboration.

Rodrigo Tavares, Invited Full Professor, NOVA School of Business and Economics (Nova SBE)

Demand for environmental, social, and governance (ESG) data has grown exponentially over the last decade, but this expansion has also generated measurement noise, weak comparability, and contested interpretations of what ESG is for. This paper argues that ESG integration becomes financially defensible only when it is governed by financial materiality, so that the metrics used in valuation, underwriting, and portfolio decisions have a credible linkage to cash flows, access to finance, or cost of capital. The paper shows that separating enterprise value disclosure from impact/ethical narratives improves decision-usefulness, and it sets out the resulting implications for governance, internal controls, reporting design, investor modelling, and stewardship. For executives, the operating implication is simple, govern material ESG like financial reporting, by selecting a small set of decision-critical ESG indicators, and linking them explicitly to capital allocation, financing strategy, and incentives.

Itai Agur, Senior Economist, Macro-Financial Division, Research Department, International Monetary Fund (IMF)
Germán Villegas-Bauer, Economist, Macro-Financial Division, Research Department, International Monetary Fund (IMF)
Tommaso Mancini-Griffoli, Assistant Director and Chief of Payments, Currencies, and Infrastructure Division, Monetary and Capital Markets Department, International Monetary Fund (IMF)
Maria Soledad Martinez Peria, Assistant Director and Chief of Macro-Financial Division, Research Department, International Monetary Fund (IMF)
Brandon Joel Tan, Economist, Payments, Currencies, and Infrastructure Division, Monetary and Capital Markets Department, International Monetary Fund (IMF)

This paper provides a conceptual framework centered on market inefficiencies to investigate the potential implications of the tokenization of financial markets. Tokenization is the creation of assets or asset representations on a digital ledger that is shared among, and trusted by, market participants and that enables self-executing contracts to be programmed onto it. Such a ledger could retire asset record keeping services, reducing costs related to asset issuance, servicing, and redemption. Tokenization may also decrease asset trading costs by addressing certain forms of counterparty risk more efficiently, speeding up settlement cycles, and lowering search frictions on specialized markets. If tokens are on-ledger assets rather than asset representations, asset custodianship costs could be saved too. Tokenization can also affect shock transmission externalities by influencing financial institutions’ funding liquidity and leveraging incentives, increasing interconnectedness, and automating trading. Potential impacts on externalities arising from networks, innovation, and market infrastructure are discussed as well. Additionally, shared ledgers can lower broker-switching costs, harnessing competition to lower investors’ fees. However, if tokenization facilitates direct trading without brokers, this could amplify internalities from limited investor understanding of risks.

Mohammad Asif Ali, Technical Lead, PNC Financial Services Group, Inc.

Loan underwriting remains a critical yet resource-intensive function within financial services institutions, often constrained by manual workflows, rigid rule-based systems, and limited adaptability to evolving risk factors. This study investigates the application of an agentic artificial intelligence (AI) architecture for end-to-end underwriting automation, combining large language models (LLMs), retrieval-augmented reasoning (RAG), and robotic process automation (RPA). The proposed framework enables autonomous data ingestion, contextual policy evaluation, and dynamic decision support while maintaining human oversight at key control points. Experimental observations indicate meaningful improvements in processing efficiency, decision consistency, and operational scalability when compared to conventional underwriting approaches. The paper also discusses governance considerations, including explainability, regulatory alignment, and bias mitigation, positioning agentic AI as a practical and extensible foundation for next-generation underwriting systems.

Udo Milkau, Digital Counsellor, Frankfurt, Germany

The vision of “autonomous agents” powered by Artificial Intelligence (AI) is far from new, beginning in parallel with the early advances in AI in the 1950/60s and has earlier roots in science fiction literature. At present, there is a lot of hype around “agentic AI” based on large language models (LMM), which are believed to be able to plan, understand, or even reason. However, all LLMs are statistic tools for estimating a “next best token” for a given prompt, typically accompanied with an error, which in turn accumulates for longer runs even for small error probabilities. The contemporary concept of AI agents is a ruled-based extension of LMMs with pre- and postprocessing, which extend the fundamental capabilities of LLMs as “generators of fiction”. However, recent claims that such systems could rewrite the rules of business and banking are based more on speculation than quantitative results. Implementations in industries such as banking cannot avoid the generic limitations of such statistical tools and have to be evaluated based on their statistical quality and improvements in performance as compared to existing process automation.

Chris A Richardson, Founder, PerceptionNexus Analytics LLC., and Executive-in-Residence, 
Department of Finance, University of Miami, Herbert Business School

Digital infrastructure systems, particularly those related to instant payments, are crucial for modern economies as they encourage faster and more comprehensive financial integration and facilitate economic growth. In contrast to physical infrastructure, digital infrastructure benefits from network effects and scalability, making it easier to include or exclude users and implement varied pricing structures. This article illustrates how the theory of public goods can help frame decisions regarding digital infrastructure and in particular digital payments. Public goods theory suggests that, in many cases, the public financing of digital payments infrastructure can lead to higher social welfare due to increased access and usage, driven by positive network externalities.

Christoph Künzle, Senior Lecturer, ZHAW School of Management and Law
Claude Baumann, Executive Chairman, FIN21 AG
Sarina Feldmann, Research Associate, FIN21 AG

Switzerland and Liechtenstein remain among the most sophisticated global wealth management hubs [Lagassé et al. (2024)]. Yet the quantitative evidence for 2025 reveals widening structural fractures beneath favorable market-driven increases in assets under management (AUM). Organic growth has stagnated, net new money (NNM) has contracted, profitability has slightly improved despite positive financial markets, and operational efficiency remains inconsistent. Rising personnel costs, technological fragmentation, and uneven governance structures further amplify strategic vulnerabilities. This paper analyses empirical performance patterns across growth, profitability, efficiency, and capital adequacy, interprets their structural implications, and formulates strategic recommendations to support long-term competitiveness.

Stephen Peters, Founder, Averus.ai
Guy Moons, Founder, Deeguisa Consulting

Transaction banking faces structural transformation as artificial intelligence (AI) and programmable settlement technologies converge. This article examines how leading institutions deploy production AI systems through integrated platforms spanning cash management, trade finance, and supply chain finance. Drawing on implementations at JPMorgan and DBS, alongside regulatory developments including MiCA and the U.S. GENIUS Act, we demonstrate measurable outcomes: 50% reduction in manual exceptions, 90% reduction in compliance false positives, and settlement compression from days to seconds. We argue that data network effects and platform economics will drive industry consolidation, as leading institutions accumulate advantages that mid-tier banks cannot replicate. The article concludes candidly: most banks have only begun the AI journey. Significant challenges remain in governance, workforce transition, and vendor readiness. Competitive advantage will accrue to institutions treating AI as strategic infrastructure requiring integrated platforms and organizational restructuring around client outcomes.