Risk Taxonomy as Governance Infrastructure: Adaptation, Traceability and Industry-led Use Cases for Fintech Innovation
Risk taxonomies in financial services are often presented as stable classification models: a hierarchy of principal risks that supports aggregation, governance oversight, and regulatory reporting. This paper argues that such a view is incomplete. In practice, a risk taxonomy is a governance infrastructure shared across stakeholder communities, including firms, supervisors, regulators, and the risk profession. Its value depends on traceability.
By traceability, we mean the practical ability to track and explain how risk categories, data, judgments, mapping rules, and mitigations are defined, applied, changed, and communicated within and beyond an organisation. Traceability matters because the quality of risk management is rarely observable in real time. Weaknesses may only become visible much later as business performance outcomes, operational incidents, or supervisory concerns, at which point feedback is costly, and remediation may be too late for firms and the wider system.
For CFO and related decision-makers, the central issue is not whether a taxonomy contains the right high-level categories (which tend to be stable across major firms), but whether the organisation can operate the taxonomy reliably: govern changes, manage mappings and aggregation, and produce assurance-grade evidence that connects categories to decisions and outcomes
Conceptually, we position risk taxonomy as a boundary object: a shared artefact that different communities can use for their own purposes while recognising it as “the same thing.” A taxonomy must be robust enough to maintain comparability across firms and reporting regimes, yet flexible enough to adapt to local realities: portfolio differences, organisational structure, data maturity, and changes in technology and competition. This predictable tension does not imply failure; it highlights that the taxonomy is doing coordination work across communities. The design challenge is to make this coordination auditable.
Empirically, we review risk management disclosures in the annual reports of three large UK banks over 2022–2024. We code for risk categories, implied hierarchies, grouping logic, and signals of year-on-year change. We observe strong stability at the top level (credit, market, liquidity/capital, operational, conduct, financial crime, model risk and climate-related risk as recurring themes), with change occurring mainly through shifting emphasis, greater granularity, and increased attention to non-financial and resilience-related risks. The implication is that innovation opportunities are less about reinventing categories and more about strengthening the socio-technical system around the taxonomy: data lineage, mapping logic, change control, governance workflows, assurance, and explainability, including for data-sparse or emerging risks where judgement and scenario processes play a larger role.
The paper’s contribution is a proposal for industry-led use cases, as a portfolio of well-specified problem statements that large financial services firms can publish to invite targeted innovation from fintechs. Each use case is designed to be procurement-ready and assurance-aware. The portfolio is organised by decision ownership, minimum data inputs, workflow controls (audit trail, segregation of duties, approvals), outputs (MI, reporting support, evidence packs), and success measures (time and cost saved, reduction in reconciliation burden, fewer classification disputes, improved supervisory confidence).
We conclude with implications for the UK supervisor. Rather than advocating new rules, we propose that supervisory engagement can be strengthened by encouraging firms to set expectations for traceability-by-design around taxonomies. By this, we mean controlled change, transparent mapping between internal granularity and external reporting, auditable lineage from source data to risk decisions, and explicit governance of data-sparse risks., including scenarios, assumptions logs, and review cycles. These steps can reduce reporting friction while improving the credibility of risk disclosures and the resilience of firms and the system.
About FRIL
Agentic AI for Scaling Targeted Support: A Governance Framework for the FCA Advice–Guidance Boundary
The Advice–Guidance Boundary Review (AGBR) introduces targeted support as a new regulated activity intended to address the persistent financial advice gap in the UK. While generative AI technologies offer the potential to scale accessible financial support, doing so within the advice–guidance boundary introduces significant governance challenges. Compliance requires structural control over segmentation logic, boundary monitoring, knowledge governance, vulnerability detection, and audit transparency.
This white paper proposes an agentic AI governance framework that embeds these regulatory functions within the architecture of AI-enabled financial support systems. The framework distributes responsibility across specialised agents responsible for segmentation governance, boundary monitoring, vulnerability detection, knowledge management, and supervisory audit. By embedding compliance functions as interacting agents surrounding a multimodal generative AI interface, the proposed architecture transforms regulatory compliance from a behavioural expectation into a structural property of the system. The framework provides a conceptual foundation for scaling targeted pensions support safely and transparently under the FCA’s AGBR while supporting responsible innovation in AI-enabled financial services.
About FRIL
Building Pre-Commercial Challenge Spaces: Innovation Calls, Regulatory Adaptation and Productivity Pathways in Financial Services
This paper examines how challenge-led innovation calls can help shape and guide the emergence of a business and policy ecosystem around current regulatory problems in financial services. Drawing on field materials from the first three Financial Regulation Innovation Lab (FRIL) innovation calls held in 2024, it argues that relatively small publicly supported challenge programmes can identify promising firms. As importantly, they can create structured pre-commercial spaces in which financial services companies, fintechs, and intermediaries can articulate common problems, test emerging solutions, and develop pathways towards pilots and adoption.
The paper is situated in a wider policy context in which productivity, innovation and sector growth have again become central concerns in UK policy. The UK’s Modern Industrial Strategy identifies Financial Services and Digital and Technologies among priority sectors for long-term growth, while recent policy developments around procurement innovation show growing interest in reducing barriers between innovative smaller firms and large institutional end users (UK Government, 2025a). The Productivity Institute, meanwhile, has argued for a broader understanding of productivity centred not only on output-input ratios but also on coordination, knowledge use, capability development and diffusion (Coyle, 2021; Jones, 2023; UK Government, 2025b; van Ark, de Vries and Pilat, 2024).
The paper treats FRIL as an intervention in productivity-relevant processes. Across the first three calls, the immediate challenges for financial-services firms were compliance-related: AI-enabled compliance simplification, ESG and sustainability compliance, and Consumer Duty compliance. Yet the practical work extended beyond compliance into data, reporting, governance, customer processes, and operational change. The first call was structured around sponsor-specific use cases, provided strong lead-user access, and tended towards concentrated matching within particular firms. Later calls, especially the ESG-focused second call, moved towards co-developed challenge statements and multi-sponsor backing, creating stronger conditions for shared learning, diffusion, and wider ecosystem formation.
It is too early to claim direct measured productivity gains at the programme level. However, the evidence from the three calls is sufficient to identify several plausible productivity pathways: clearer problem articulation, reduced search and matching frictions, faster pre-commercial validation, reduced manual effort in compliance-related processes, and stronger conditions for diffusion and reuse. The central policy lesson is that innovation calls work best when they are treated both as competitions for technical ideas, and as mechanisms for shaping and guiding pre-commercial ecosystems around current business and policy problems.
About FRIL
AI Governance after MiFID II: Beyond (Mere) Technological Neutrality?
This article examines the evolving intersections between artificial intelligence (AI) and EU financial regulation, focusing on the Markets in Financial Instruments Directive II (MiFID II). Grounded in the principle of technological neutrality, MiFID II seeks to enhance investor protection, safeguard market integrity, and ensure that innovation develops within competitive and well-regulated markets across the Union. The article argues, however, that while this neutrality renders the framework functionally enabling, it also leaves it normatively silent in the face of the distinctive and evolving risks introduced by financial AI. As AI applications become increasingly heterogeneous—both across the financial functions in which they are deployed and in their underlying lifecycles and value chains—MiFID II’s activity-based logic increasingly struggles to accommodate their diverse and evolving risk profiles. Reflecting the EU’s broader shift toward risk-based AI governance, the article outlines an initial taxonomy of financial AI applications designed to guide the proportionate alignment of regulatory obligations with AI-related risks, thereby supporting the continued adaptability, coherence, and future-proofing of EU financial services law.
About FRIL
Multimodal AI for Scaling Targeted Support: Navigating the FCA Advice–Guidance Boundary
Summary
The Financial Conduct Authority’s Advice Guidance Boundary Review (AGBR) seeks to address a persistent advice gap in UK financial services by enabling new forms of scalable, decision-relevant consumer support that sit between generic guidance and personalised financial advice. This challenge is particularly acute in pensions, where consumers face complex, long-term decisions but exhibit low engagement with traditional advice services. This white paper examines the potential of multimodal generative artificial intelligence to deliver targeted support in pensions while remaining within the advice–guidance boundary. Drawing on recent advances in Vision Language Models and multimodal conversational architectures, the paper develops a solution framework for speech-enabled, audio-visual digital advisors that are compliant by design. A Digital Pensions Advisor prototype is presented to demonstrate how such systems can interpret real consumer narratives, recognise and respond appropriately to vulnerability, and maintain boundary discipline when confronted with requests for personalised advice. The paper concludes by outlining a roadmap for strengthening auditability, explainability, and supervisory readiness, and by identifying future research directions, including the role of formal and informal information sources in shaping consumer understanding. Collectively, the findings suggest that multimodal AI can play a meaningful role in scaling targeted support in pensions while preserving regulatory safeguards and consumer trust.
Hao Zhang Digital Avatar demo
About FRIL
Consumers at the Heart of Innovation: Financial Health Evaluation in the UK Regulatory Landscape
This whitepaper examines the evolving landscape of financial services in the UK, with a focus on consumer-centric innovation and the regulatory framework supporting financial health. Against a backdrop of rising household debt and consumer vulnerabilities, we explore how innovative fintech solutions, enabled by emerging technologies, are reshaping the evaluation of financial health beyond traditional credit scoring methods.
The paper provides an in-depth analysis of the current regulatory environment, including the Financial Conduct Authority’s (FCA) Consumer Duty and other initiatives aimed at enhancing financial wellbeing. A case study demonstrates an AI-driven framework for intelligent credit scoring, illustrating the potential for more holistic financial health assessments.
We discuss the critical role of Open Banking and Open Finance as enabling infrastructures for innovation, while emphasising the importance of data protection and digital ethics. The paper also outlines key directions for fintech innovation that prioritise consumer needs.
Looking to the future, we examine emerging trends in consumer innovation, potential regulatory developments, and the long-term impact on the financial services industry. The paper concludes that by placing consumers at the heart of innovation, the UK financial sector can build a more resilient, inclusive, and sustainable future, setting a global benchmark for fostering financial health in an era of rapid technological change.
About FRIL
From Crisis to Prosperity: AI and Open Finance for Holistic Financial Health and Smart Future Planning
The financial services industry stands at a critical inflection point. Traditional credit-scorecentric frameworks, rooted in historical repayment data and broad demographic categories, are increasingly incapable of capturing the full spectrum of consumers’ financial health. The COVID-19 pandemic, successive cost-of-living crises and wider economic shocks have exposed deep vulnerabilities in reactive, one-dimensional risk models. Consumers and institutions require proactive, resilience-focused insights that span day-to-day cashflow management, medium-term debt servicing, and long-term wealth accumulation.
Open Finance – the consent-driven sharing of a comprehensive array of financial data (current accounts, mortgages, loans, savings, investments, pensions, insurance, government benefits), combined with advanced Artificial Intelligence (AI) and Machine Learning (ML) techniques, offers a transformational route to truly holistic financial health evaluation. By ingesting rich, multi-dimensional data and applying explainable models, financial firms can transition from static credit assessments to dynamic, personalised guidance and recommendation engines. These engines empower consumers to build financial resilience, make informed decisions and pursue life goals with confidence.
Building on Sopra Steria and Glasgow University’s joint whitepaper “Consumers at the Heart of Innovation: Financial Health Evaluation in the UK Regulatory Landscape” and inspired by R&D collaboration between Sopra Steria and Oxford University, this whitepaper:
- Examines the scope and expected data architecture of Open Finance, extending from Open Banking to full-spectrum data sharing.
- Presents a robust data modelling framework, encompassing data acquisition, cleansing, feature engineering, supervised and unsupervised modelling, scorecard design and persona segmentation. It culminates in a composite financial health score that blends aspects like credit risk and resilience evaluation.
- Explores explainability and consumer engagement, employing data science techniques to ensure transparency and mapping various persona archetypes to tailored, sequential optimisation plans.
- Demonstrates regulatory alignment, showing how non-product-specific, personadriven guidance and recommendation can fit within the Financial Conduct Authority’s (FCA) Advice and Guidance Boundary (FG15/1) and the Consumer Duty framework.
About FRIL
Supply Chain Intelligence: Actionable Risk Assessment of Brazilian Commodity Supply Chains Using Geospatial Data
Geospatial data is transforming sustainability risk assessment in Financial Services. Driven by mandatory regulations such as the EU’s CSRD and SFDR, alongside emerging standards such as TNFD, financial institutions must now monitor not just financial performance, but real-world environmental and social impacts across global supply chains.
This white paper showcases how geospatial data and asset-level geospatial analysis can be used as a support tool for supply chain impact monitoring. Specifically, we evaluate deforestation risks in Brazilian commodity supply chains. Using publicly available datasets and Google Earth Engine, we develop a reproducible risk scoring framework, applied to over 17,000 slaughterhouse facilities, tied to 9,854 companies, across Brazil.
This analysis:
- Quantifies deforestation exposure across 12 animal-based commodities at facility and company level.
- Creates actionable risk metrics for investors, lenders, and regulators.
- Aligns outputs with ESG disclosure frameworks, including CSRD, SFDR, TNFD, and the EUDR.
- Highlights high-risk companies and regions, providing clear signals for due diligence and sustainable finance strategy.
This paper is part of a broader move toward Earth Intelligent Finance, empowering financial actors to make faster, smarter, and more transparent sustainability decisions using geospatial insight.
Mind the Gap: Bridging the UK’s pension divide with digital solutions
The UK pensions landscape is at a crossroads: with 38% of the working-age population under-saving for retirement and 52% of people accessing their pensions without adequate advice, the risk of poor financial outcomes are at an all-time high. Yet, the market is set for huge growth, with defined contribution pension assets projected to explode to £800 billion by 2030.
This is not a new problem, and despite well-intentioned efforts like auto-enrollment from policy-makers and new digital products from providers, there is still a fundamental disconnect; customers lack engagement and understanding in later life planning and financial outcomes.
Solving this challenge presents a significant growth opportunity for providers, but to do so needs a design-led approach focused on deeply understanding your customers through three core elements: Empathy, Engagement and Empowerment. Technology – and particularly AI – of course plays a crucial role, but doesn’t replace the need for empowered advisers. The future will be hybrid: a combination of traditional digital interfaces, agentic AI, and human touchpoints to create hyper-personalised experiences based on an individual’s aspirations, circumstances and preferences.
This whitepaper, drawing on digital partner CreateFuture‘s decades-long experience in the Wealth & Pensions sector, provides a compelling vision for the future of pensions engagement, outlining how a hybrid approach leveraging digital innovation, AI, and empowered advisors can create truly customer-centric experiences and unlock a more secure future for savers.
Large Language Model Application for Regulatory Horizon Scanning: Case Study on ESG Greenwashing Regulations
This white paper explores the application of Generative AI, specifically Large Language Models (LLMs), to enhance regulatory horizon scanning within financial services. Using the Financial Conduct Authority’s (FCA) 2024 anti-greenwashing rule as a case study, we demonstrate how LLMs can be integrated into the strategic foresight process to detect early regulatory signals, analyse stakeholder feedback, and forecast future regulatory developments.
Our framework builds upon the traditional horizon scanning model, comprising exploration, assessment, application, and continuation, and incorporates advanced text analysis techniques including semantic similarity testing with models such as BERT and RoBERTa.
The study shows that LLMs can significantly improve the efficiency, accuracy, and scalability of horizon scanning by extracting meaningful insights from large, unstructured datasets. The results highlight the potential of LLM-driven foresight to enhance regulatory preparedness, guide compliance strategies, and inform policy design in an increasingly complex and dynamic regulatory environment.